The No Child Left Behind Act of 2001 (NCLB) directs the Department of Education to distribute Title I basic and concentration grants directly to school districts on the basis of the most recent estimates of children in poverty available from the Census Bureau. These estimates are produced under the Census Bureau's Small Area Income and Poverty Estimates (SAIPE) program. The estimates are based on Census 2000 and the SAIPE program's model-based estimates of poverty for all counties.

The 2001, 2002, and 2003 estimates are for school districts according to their 2003-2004 boundaries.

Three estimates are provided for each school district:

• total population
• children ages 5-17
• related children ages 5-17 in families in poverty

The number of related school-age children in families in poverty in each school district is provided as a component of the determination of Title I grants. The estimate of the total population of each district is provided for use in the small district (less than 20,000 population) provision. The figure for school-age children is provided so that the proportion of children in poverty can be determined. This proportion is required for determining eligibility for grants. A true poverty "rate" for children cannot be determined from these figures, because the numerator and denominator refer to slightly different universes.

The school district estimates are based upon tabulations of poverty from Census 2000, using school district boundaries corresponding to school year 2003 - 2004. These tabulated data are then combined with the SAIPE program county estimates of poverty, utilizing methods consistent with the SAIPE program school district estimates since income year 1995. By construction, the SAIPE program school district estimates are arithmetically consistent with the SAIPE program county and state estimates, and with the national estimate from the Annual Social and Economic Supplement (ASEC) of the Current Population Survey (CPS) - the official source of national poverty estimates.

We provide poverty estimates for all school districts that are in the Census Bureau's TIGER database, updated by each School District Boundary project; the most recent was completed in 2004. We also tabulate and produce estimates for all occupied areas not assigned to any school district. We refer to these areas as "balances" of the counties in which they occur, whether they comprise a single compact area or not. Although we do not publish estimates for "balance of county" areas on our website, they are provided to the U.S. Department of Education for implementing provisions of NCLB, and are available upon request.

For each school district, our estimates pertain to all resident school-age children ages 5-17, inclusive, whether enrolled in public or private school, or not enrolled. Where two districts divide the children of an area between them by grade, the estimates do so as well. In most areas, districts called "elementary" or "unified" are, no matter their names, responsible for providing education for all elementary and secondary grades - either by operating schools themselves or by purchasing instruction from neighboring school districts - for all residents of their territory. In these areas, data for all people ages 5-17, inclusive, are tabulated in the district in which they reside.

Some states have areas with separate "elementary" and "secondary" school districts, each exclusively responsible for providing education in some grades in their shared territory.¹ In these areas, data for school-age children are allocated between districts on the basis of the grade range of the district and the grade assigned to the child. There are also some states that have school districts with different grade ranges in different parts of their territory.² In most cases these are districts that are "unified" in part of their domain, and "secondary" in the rest. The final tabulations and estimates reflect the combination of data honoring these distinctions.

Grade ranges for each district are collected during the boundary update, and supplemented with phone calls to districts. We attempt to assign a single grade range to each district which, in the case of spatially overlapping districts, leaves no grade unclaimed and no grade claimed by more than one district. Occasionally the pattern of grade ranges of overlapping districts does not permit each grade to be assigned to exactly one and only one district. In these few instances, three simple rules are applied:

• if a "unified" district is present, data for children whose assigned grade is claimed by two districts or not claimed by either are allocated to the "unified" district
• if an "elementary" and "secondary" district are present, data for children whose assigned grade is claimed by both are allocated to the "secondary" district
• if an "elementary" and "secondary" district are present, data for children whose assigned grade is claimed by neither are allocated to the "elementary" district.

To tabulate the data for each district, each child is assigned a grade. In the Census 2000 sample, where responses to the "long-form" questions are available, 97 percent of children are assigned a grade on the basis of their edited reports of the grade in which they were enrolled. Because this question used response categories that represent multiple grades (PK, KG, 1st - 4th, 5th - 8th, 9th-12th, higher), the child's age in October was used to assign single grades from among those implied by the answer. For those not enrolled, the modal grade for their age in October (age on October 1, 1999 less 5) was assigned, provided that the grade assigned was not reported as having been completed. For Census 2000 short-form data, where school enrollment and educational attainment are not available, children were assigned the modal grade for their age on October 1, 1999.

With the Census 2000 record for each child assigned to a single 2003-2004 school district, to which that child is said to be "relevant," we tabulate for each district:

• "relevant" children ages 5-17, inclusive
  
• "relevant" children who are related to the householder and reside in families whose 1999 income falls below the poverty threshold appropriate to that family
• total population of all ages residing in the spatial boundaries of the district. (Note that because of overlapping districts, the sum of the total populations of the districts will exceed the total population of the nation.)

Related children are people ages 5-17 related by birth, marriage, or adoption to the householder of the housing unit in which they reside; foster children, other unrelated individuals, and residents of group quarters are not "related children".

The SAIPE program procedure for estimating poverty among relevant children ages 5-17 in families works with geographical units we call school-district-county-pieces. These pieces are defined as the intersections of school districts and counties (i.e., all of a district if it does not cross county boundaries and each county part separately for districts that do). If a school district has territory in two counties, for example, we make estimates for the two parts separately and then combine them. School districts with territory in a single county are composed of a single piece.

The first step in making the school district poverty estimates is to estimate poverty rates for income year 1999 from Census 2000. The poverty rates for relevant school-aged children in families are estimated using a method called "estimated best linear unbiased predictor" (EBLUP) with a simple model. Using EBLUP, the estimate for each school district piece in a county is the weighted average of the direct sample estimate (produced as described above) for all of the school district pieces in that county. The weight for a school district piece depends on the relative variance of its sampling error and the variance between the school district pieces' true poverty rates within the county. The effect of this procedure is to "shrink" the estimates toward a county-wide poverty rate, so we often refer to this kind of estimator as a "shrinkage" estimator.

School district pieces can be very small, in which case the sampling-error variances of the estimates from Census 2000 sample data can be very high. In extreme cases, the Census 2000 estimates of the number of people in poverty in a school district might be zero or may exceed the population. Since these estimates are used throughout the decade, the effects of large sampling errors can have lasting effects. The shrinkage procedure produces estimates of poverty rates that are greater than zero and less than 100 percent in nearly every case. Further, the overall magnitude of the error is reduced under the fairly general model for which these estimates are derived. Beginning with the 2003 estimates, we implemented an improved method for our shrinkage estimator. The expected value of poverty rates for school district pieces within counties are unchanged, but the standard errors are reduced. This results in small changes for most school districts, but some districts may have larger changes.

To get estimates of the number of relevant children ages 5-17 in families in poverty , we multiply the shrinkage estimate of poverty rates by the Census 2000 counts of the numbers of relevant related children ages 5-17. The numbers of children in poverty are then adjusted, using "controlled rounding," to get a result with the following properties:

• The number of children in poverty in the pieces in a county adds up to the SAIPE program's model-based estimate of the number of children in poverty in their counties for the relevant income year.
• The number of children in poverty in the school district pieces are integers.

This is possible, in part, because the county-level estimates have already been adjusted so they have the same properties relative to the states, which, in turn, have the same properties relative to the national CPS ASEC estimated number of school-aged children in poverty.

The controlled rounding has the approximate effect of calculating the shares of a county's children in poverty who reside in the school district pieces from the shrinkage estimates and apportioning the SAIPE program county estimated children in poverty to the school district pieces according to those shares. The final step is to reassemble the school district pieces into the school districts, simply by adding their controlled-rounded numbers of children in poverty together.

¹States where districts may overlap: Arizona, California, Connecticut, Illinois, Maine, Massachusetts, Montana, Nebraska, New Hampshire, New Jersey, New York, Oregon, Rhode Island, South Carolina, Tennessee, Vermont, and Wisconsin.

²States where grade ranges may differ within a district: Massachusetts, Nebraska, Oregon, South Carolina, Tennessee.