The 2001 Current Population Survey (CPS) served as a tool for testing a sample expansion of the October School Enrollment Supplement and as a bridge to introduce new Census 2000-based population controls. The following section discusses the effects these methodological changes had on school estimates.

The Census Bureau tested a 9,000 household expansion in the interviewed sample for the CPS October School Enrollment Supplement in 2001. The original sample size of approximately 47,000 interviewed households for the 2001 CPS School Enrollment Supplement was increased to approximately 56,000. The primary goal of the sample expansion was to produce more reliable estimates of low-income children without health insurance for the State Children’s Health Insurance Program (SCHIP) through reduced variances. Although the SCHIP sample expansion was specifically targeted toward producing better children’s health insurance estimates at the state level, other state estimates, as well as national estimates, improved. Further information about the SCHIP sample expansion is available on the internet at https://www.bls.gov/cps/cpsjul2001.pdf.

The procedure used in developing estimates for the entire civilian noninstitutional population for the Current Population Survey (CPS) involves the weighting of sample results to independent estimates of the population by sex, age, race, and Hispanic/non-Hispanic categories. These independent estimates are developed by using civilian noninstitutional population counts from the decennial censuses and projecting them forward to current years using data on births, deaths, and net migration. Beginning with the 2001 CPS October School Enrollment Supplement, the independent estimates used as control totals for the CPS are based on civilian noninstitutional population benchmarks consistent with Census 2000.

Table C-1 displays national-level data for school enrollment from the original and expanded CPS samples from October 2000 and October 2001. The introduction of the expanded sample for October 2001 had very little effect on school enrollment. Changes that were statistically significant are shown in Table C-1, column 5.¹ Among the population three years and over, there was a statistically significant increase (0.2 percentage points) in the percent enrolled in college. This was the only statistically significant change attributable to the expansion of the sample.

Weighting the estimates with 2000 population controls, instead of the 1990 census controls used in previous reports, had very little effect on school enrollment in 2001 (see Table C-1, column 7). The proportion of the population age three years and over enrolled in school decreased. The only statistically significant increase was for the proportion of the population three years and over who were not enrolled in school (0.4 percentage points).

For further information about CPS weighting procedures, see Technical Paper 63RV.

1. Usually when two estimates are "significantly different" it means that the difference was large enough, in relation to the difference’s own standard error, for us to infer that the difference is "real," or more accurately, that there was a less than 10 percent chance that the difference merely came from sampling variation. In Table C-1, however, both sets of data estimate the same populations in the same period. Therefore, "significant" here means that we would have inferred that the estimates came from different populations, if we did not already know they were the same.