Screening interviews were completed with some 63,844 households in 1993, some 55,838 households in 1996, some 55,929 households in 1999, some 32,049 households in 2003, and some 54,034 households in 2007. The unit response rate for the screener interview in each of these five survey years was at 82 percent in 1993, at 70 percent in 1996, at 74 percent in 1999, at 65 percent in 2003, and at 53 percent in 2007.
The unit response rates for the extended interview were 90 percent for the 1993 School Readiness Survey (ages 3 through 7 or in 2nd grade or below), 89 and 90 percent for the 1993 School Safety and Discipline Survey (3rd through 5th grade, and 6th through 12th grade, respectively),1 89 percent for the 1996 Parent and Family Involvement in Education Survey, 88 percent for the 1999 Parent Survey, 83 percent for the 2003 Parent and Family Involvement in Education Survey, and 74 percent for the 2007 Parent and Family Involvement in Education Survey.
The overall unit response rates (the product of the screener response rate and the extended interview response rate) were 74 percent for the 1993 School Readiness Survey (ages 3 through 7 or up to 2nd grade), 73 and 74 percent for the 1993 School Safety and Discipline Survey (3rd through 5th grade and 6th through 12th grade, respectively), 63 percent for the 1996 Parent and Family Involvement in Education Survey, 65 percent for the 1999 Parent Survey, 54 percent for the 2003 Parent and Family Involvement in Education Survey, and 39 percent for the 2007 Parent and Family Involvement in Education Survey.
Bias analyses have been conducted periodically on the NHES to track potential bias due to declining response rates and undercoverage, and comparisons of estimates using surveys with higher response rates have been performed for each survey year. Bias examined in the NHES has been shown to be corrected with the weighting adjustments, although such adjustments may ignore correlated bias in other variables that cannot be examined. Therefore, the potential for bias remains. Statistical adjustments used in weighting were similar across all administrations of the NHES. Detailed nonresponse bias analyses were conducted on the NHES in 1999 and 2007 and neither found substantive biases in the NHES estimates (Van de Kerckhove et al. 2009, Montaquila et al. 2008, Nolin et al. 2000). In the 2007 study, some variables related to the preschool population were shown to have potential for bias, however, those variables were not used in this report (Van de Kerckhove et al. 2009). Undercoverage bias was assessed in 1993 and 1996 and again was shown to be corrected with weights. Poor households and rented households showed potential for bias before nonresponse adjustments were applied to the base weights (Brick et al. 1997, Montaquila et al. 1997). An unpublished comparison of estimates for 2003 shows that NHES estimates of the number of students in private school was 8 percent higher than estimates in the 2001 Current Population Survey (CPS) and that NHES had 5 to 6 percent more students in modal grades compared with the 2001 CPS.
In all five survey years, item nonresponse (the failure to complete some items in an otherwise completed interview) was very low (less than 2 percent for most variables in this report). For information about specific item response rates, see the data file user's manual for each survey year. All NHES items with missing responses (i.e., don't know, refused, or not ascertained), except those which were derived from the Common Core of Data (CCD) and the Private School Universe Survey (PSS) (i.e. inapplicable in CCD file or data are missing for school), were imputed using a hot-deck imputation procedure (Kalton and Kasprzyk 1986).2 Variables taken from the CCD and PSS may contain missing data. These data were not imputed on the NHES. Cases with missing CCD or PSS information for variables applicable to this report were dropped from the analysis. Less than 0.5 percent of cases had missing CCD or PSS information. For more information on the CCD visit http://nces.ed.gov/ccd. Also see Sable, Thomas, and Sietsema 2007. For more information about the PSS visit http://nces.ed.gov/surveys/pss. Also see Tourkin et al. 2008.
1 Unlike other survey years, in 1993 there
were multiple surveys based on age group, which were combined to form the analysis
in this report. Response rates for 1993 were therefore calculated separately by
2 For more information on the imputation procedures used in NHES:1993, NHES:1996, NHES:1999, NHES:2003, and NHES:2007, see the following: Brick et al. 1997; Montaquila and Brick 1997; Nolin et al. 2000; Hagedorn et al. 2004; and Hagedorn et al. 2008.