The following text provides a brief overview of the estimates of related school-age children in families in poverty for school districts for income year 1997.
The Improving America's Schools Act of 1994 directed the Department of Education to consider distributing Title I basic and concentration grants directly to school districts for the 1999-2000 school year. The Department of Education asked the U.S. Census Bureau to provide estimates at a school district level that would make this possible. The allocation is to be on a school district (local education agency) basis unless the Secretaries of Education and Commerce determine that the estimates are "inappropriate or unreliable."
That determination was aided by an evaluation of the available options by the Panel on Estimates of Poverty for Small Geographic Areas of the National Research Council. The results of this evaluation were published in 1999 as Small-Area Estimates of School-Age Children in Poverty, Interim Report 3: Evaluation of 1995 County and School District Estimates for Title I Allocations by the National Academy Press. The panel recommended that the estimates be used to make direct Title I allocations to school districts for the 1999-2000 school year.
Previously, Title I grants were distributed on a county basis with the individual states having responsibility to redistribute the funds from counties to school districts. This redistribution was done using a variety of techniques that varied by state. Many states used 1990 census data, or 1990 census data combined with data from other programs targeting a poor population, such as Aid to Families with Dependent Children, the Food Stamp Program, Medicaid, the Foster Care Program and/or the School Lunch Program. Other states used these alternative data sources only.
Under the legislation that would allocate funds directly to school districts, states could aggregate and then redistribute funds for school districts with a population less than 20,000. The National Research Council Panel recommended that states which chose this option use the county totals from the Census Bureau's SAIPE program to the extent possible.
Three estimates are provided for each school district: total population, the number of school-age children (ages 5-17), and the number of related school-age children in families in poverty. Where two districts divide the children of an area between them by grade, the estimates do so as well. The number of related school-age children in families in poverty in each school district is provided as a component of the determination of basic Title I grants. The estimate of the total population of each district is provided by the U.S. Census Bureau for use in the small district (less than 20,000) provision. The figure for school-age children is provided so the proportion of poor children can be determined. This proportion is required for allocating concentration grants. A true poverty "rate" for children cannot be determined from these figures because the numerator and denominator refer to slightly different universes.
Construction of the estimates
The poverty estimates were created using a "synthetic estimator." The estimate is the number of poor children as measured in the 1990 census (for school districts defined by geographic boundaries for the 1999-2000 school year) multiplied by the proportional change in child poverty in the county in which the district is located. The school district estimates were then adjusted so that they sum to the county estimates, which were, in turn, controlled to sum to the state estimates. The county change is computed as the change from the 1990 census to the U.S. Census Bureau's county model estimates for 1997. The county estimation models use administrative data and data from the 1990 census and the Current Population Survey to estimate child poverty.
The population figures were created using similar "synthetic estimators."
This kind of estimator can not capture the variation in changes in poverty between school districts within a county. It would be preferable to use a model with explanatory variables on the school district level, but variables that are consistent across states are not presently available.
Precision of the estimates
The "synthetic" estimates for school districts do not have the degree of precision normally associated with data published by the U.S. Census Bureau.
The U.S. Census Bureau constructed an evaluation file with the purpose of examining the precision of these estimates. It applied the estimation methods described above to the base data from the 1980 census, calculated "synthetic" estimates for 1990, and compared the results to the 1990 census. Districts were excluded from the evaluation file if they were coterminous with counties, did not cover the entire grade range, or were known to have changed geographic boundaries between 1980 and 1990. The remaining districts represented 61 percent of the total number of districts and contained 56 percent of school-age children.
The estimates proved reasonably accurate for large districts (with a population greater than 40,000) and for those coterminous with counties. These districts comprise 13 percent of districts but include 62 percent of poor children. However, there are potentially very large errors for small districts. The average absolute difference between the estimate of related school-age children in families in poverty and the same figure from the 1990 census is 60 percent of the census figure for school districts in the evaluation file. The results for large districts are similar to the county model results, which have a comparable figure of only 16 percent.
The estimates for the number of school-age children and the total population are relatively more accurate. The estimate of the number of school-age children in the evaluation file differs from the 1990 census figure by 17 percent on average. For the estimate of total population, the difference is only 13 percent.
While the error relevant to the average school district is large, the error relevant to the average poor child is much less. The average absolute difference between the evaluation file estimate and the census figure is only 22 percent of the average number of poor children per district. This result is obtained because most children are in large districts where the estimates are more accurate.
Recommendations of the Panel on Estimates of Poverty for Small Geographic Areas, National Research Council, National Academy of Sciences
Despite the limitations, the panel recommended to the Secretaries of Education and Commerce that the Department of Education use the school district estimates for FY 1999 Title I allocations. Even though the potential errors are large, the panel does not view the estimates to be "inappropriate or unreliable". The panel interprets this clause in a relative sense: the estimates are "generally as good as--and, in some instances, better than--estimates that are currently being used." (1) As an example of such a comparison, the panel compared the school district estimates to estimates incorporating school lunch data for the state of New York. The two sets of results differed only marginally.
The report mentions other factors that were also important in the Panel's decision. The Congressional mandate for direct allocations to school districts had a heavy weight, as well as the uniformity gained by using the U.S. Census Bureau's school district estimates. Also, the panel determined that the U.S. Census Bureau's ratio-adjusted estimates from the 1990 Census have a lower variability than the estimates that states are likely to be able to derive from the same source. (2)
The Secretaries of Commerce and Education have determined that these school district estimates are not "inappropriate or unreliable" and, thus, they will be used for Title I funding.
1 See Interim Report 3, page 3.
2 See Interim Report 3, page 45.