WASHINGTON--Updating poverty figures between decennial census counts to better target federal education funds would be both difficult and expensive, and estimates for small school districts could prove unreliable, statisticians told lawmakers last month.
Frequent updating of estimates at the district level “would require both data and methodology we do not now have,’' William P. Butz, the associate director for demographic programs at the Census Bureau, said at a joint hearing of the House Subcommittees on Elementary, Secondary, and Vocational Education and Census Statistics and Postal Personnel.
Data from the 1990 census are being used for the first time this year to allocate funds under federal programs whose formulas are driven by census counts, including Chapter 1 and several other education programs.
As a result of population shifts, many districts are experiencing significant gains or losses in federal funding, sparking lawmakers’ interest in legislative changes that would allow more gradual shifts based on updated population counts.
Allocations under Chapter 1, and under several other programs that use the Chapter 1 formula in whole or in part, are based primarily on census counts of children from poor families. To date, the Census Bureau has collected poverty data only at the county level, and only as part of the once-a-decade census. Annual updates provide only general population figures without additional demographic information.
Time Lag
As a result, funding is distributed on the basis of data that can be as many as 13 years old. That means districts with large influxes of poor children can receive no additional funding for them for many years.
In addition, districts where poverty declines--or does not rise as much as in other parts of the country--experience precipitous drops in funding when new data begin to be used.
In an effort to attack the problem, Rep. Tom Sawyer, D-Ohio, the chairman of the census subcommittee, introduced legislation that would require the Census Bureau to estimate biannually the number of poor, school-age children in each state, county, city, and district, beginning in 1995.
But Mr. Butz, Commissioner of Education Statistics Emerson J. Elliott, and Patricia Ruggles, a researcher at the Urban Institute, said it would be virtually impossible to produce such data accurately.
The Census Bureau is now working with the Education Department on its first district-level estimates, based on the 1990 census, Mr. Butz said, but has just begun research on how to update age and income data for areas smaller than states.
Changing and overlapping district boundaries present logistical difficulties, he said, but they pale in comparison to the potential problems associated with trying to update age and income data between census counts, when such statistics can be drawn from detailed household surveys.
The best source of updated information is tax returns, the statisticians said, but some people do not file them, and poor people are the least likely to file.
Alternative Data
Counts of families receiving welfare payments or reduced-price school lunches could be used to update “baseline’’ data from the census. But welfare-eligibility requirements vary by state, and lunch counts are influenced by the accuracy of parental reporting, parents’ willingness to fill out applications, and district policy. Ms. Ruggles said more families apply for the subsidies in districts where lunch prices are higher.
The lunch program “doesn’t collect information on income, it collects information on how many families have low income and want their children to receive free or low-priced lunches,’' she said.
In any case, Mr. Butz said, the Census Bureau’s attempts at mid-decade population estimates have proved unreliable for units smaller than 5,000 people, and 74 percent of all districts enroll fewer than 2,500 children.
He said the agency can probably produce reliable estimates for areas with at least 50,000 residents, but he is less sanguine about the prospects for more targeted updates.
Mr. Elliott concurred, “Updates of district data would be of questionable quality.’'
While district-level estimates might be reasonably accurate at a national level, he said, that is because so many poor children live in cities, where estimates are more reliable.
“Poor children in small districts are much more likely to receive too little or too much Chapter 1 money than are poor children in large districts,’' Mr. Elliott said.
The witnesses said the Census Bureau could probably update age and poverty counts for large sampling areas and use that information on population shifts to produce estimates for small districts within those larger areas.
Mr. Sawyer asked if such a method would be “adequate.’'
That “is essentially a political question,’' Mr. Elliott said.
