National Teacher and Principal Survey (NTPS)

4. SURVEY DESIGN

Target Population

Public (including charter schools) and private elementary and secondary schools in the United States; the principals of these schools; and the teachers of these schools. Charter schools are included in the study.

Private schools were dropped from the target population from the 2015–16 NTPS, due to low response rates in the 2011–12 Private School SASS and the 2013–14 Private School Survey (PSS). Private schools were included in the target population of the 2017–2018 cycle.

Sample Design

Public schools. The starting point for the 2017–18 NTPS public school sampling frame was the 2014–15 Common Core of Data (CCD) Nonfiscal School Universe data file. The sampling frame was adjusted from the CCD to fit the definition of a school eligible for NTPS. To be eligible for NTPS, a school was defined as an institution or part of an institution that provides instruction to students, has one or more teachers to provide instruction, serves students in one or more of grades 1–12 or the ungraded equivalent, and is located in one or more buildings apart from a private home. It was possible for two or more schools to share the same building; in that case, they were treated as different schools if they had different administrators (i.e., principal or school head). This definition is unchanged from the Schools and Staffing Survey (SASS). The 2017–18 NTPS universe of schools is confined to the 50 states plus the District of Columbia and excludes the other jurisdictions, Department of Defense overseas schools, and CCD schools that do not offer teacher-provided classroom instruction in grades 1–12 or the ungraded equivalent.

Unlike the SASS, the NTPS did not stratify schools prior to sampling. The NTPS used a systematic, probability proportionate to size (PPS) sample, where size is defined to be the square root of the number of full-time-equivalent (FTE) teachers in the school. Some types of schools were oversampled based on select characteristics, such as urbanicity (city, suburban, town, rural), school grade level (primary, middle, high, combined), and charter status. In addition to oversampling based on specific school characteristics, sample sizes were inflated for schools in 29 states with smaller numbers of schools, in order to ensure that all state-level estimates would meet the criteria for publish ability.

Private Schools. The 2017–18 NTPS private school frame was based on the 2015–16 Private School Universe Survey (PSS) list frame and certainty area frame. In order to provide coverage of private schools founded since 2016 and to improve coverage of private schools existing in 2016, the Census Bureau collected membership lists during the summer of 2016 from private school associations and religious denominations. The associations were asked to list all schools meeting the PSS school definition. The 50 states and the District of Columbia were also asked to provide lists of private schools meeting the PSS definition of a school.

The NTPS private school sample that was drawn from the list frame was a systematic probability proportionate to size (PPS) sample, where size is defined to be the square root of the number of full-time equivalent (FTE) teachers in the school. However, some types of schools were oversampled by being placed into domains for oversampling. The domains were defined by affiliation strata (Catholic, Baptist/Seventh Day Adventist, Lutheran/Jewish/other religious, nonreligious regular, and nonreligious special education or special emphasis), grade level (elementary, secondary, and combined), school size (large, small) and census region (Northeast, Midwest, South, and West).

Principal selection. All principals from sampled schools were also surveyed for NTPS.

Teacher selection. Prior to allocating teachers to sampling strata, the Census Bureau first allocated an overall number of teachers to be selected. The maximum number of sampled teachers per school was set at 20, in order to avoid overburdening a school by sampling too large a proportion of its teachers. An average of seven to nine teachers were selected per public school, depending on the school's grade range, school size, urbanicity, and poverty status. For private schools, an average of two to four teachers per school were selected, depending on affiliation, school size, and region. Within each sampled school (both public and private), teachers were stratified by subject, as follows: math, science, English/language arts, social studies, and everything else. No oversampling by subject was performed. Within each teacher stratum in each school, teachers were selected systematically with equal probability.

The selected samples of the 2017–18 NTPS included about 10,600 traditional and charter public schools and their principals, 60,000 public school teachers, 4,000 private schools and their principals, and 9,600 private school teachers.

Data Collection and Processing
The 2017–18 NTPS used a combination of mail-based methodology and Internet reporting for questionnaires, with telephone and in-person field follow-up. An advance letter was mailed to sampled schools during the summer of 2017 to verify school addresses and eligibility. Subsequently, a package containing school and principal surveys and explanatory information was mailed to sampled schools. The Census telephone center called sampled schools to verify school information, establish a survey coordinator, and follow up on the Teacher Listing Form (TLF), which served as the teacher list frame. Sampled teachers were mailed questionnaires on a flow basis. Field follow-up was conducted for types of schools expected to have a lower response propensity (e.g., city schools) and schools that had not returned the TLF. Schools were called from Census telephone centers to remind the survey coordinator to have staff complete and return all forms. Sampled principals and teachers were called from the telephone centers to attempt to complete the questionnaire with them over the phone. Field follow-up was conducted for schools and teachers that had not returned their questionnaires.

Data processing. The U.S. Census Bureau conducted the data processing for the 2017–18 NTPS. Each questionnaire was coded according to its response status-for example, whether the questionnaire contained a completed interview, a respondent refused to complete it, or a school closed. The next step was to make a preliminary determination of each case's interview status, i.e., whether it was an interview, a non-interview, or if the respondent was ineligible for the survey.

Once the data were compiled, a computer program conducted a series of quality control checks, such as range checks, consistency edits, and blanking edits, and generated a list of cases where problems occurred in each survey. After the completion of these checks, the program made a final determination of whether the case was eligible for the survey, and if so, whether there were sufficient data for the case to be classified as an interview. As a result, a final interview status recode value was assigned to each case.

Estimation Methods
Sample units are weighted to produce national and state-level estimates for public and private elementary and secondary school surveys (i.e., schools, teachers, and principals). These estimates are produced through the weighting and imputation procedures discussed below.

Weighting. Estimates from NTPS sample data are produced by using weights. The base weight for schools and principals is the reciprocal of the probability of selection for each school and principal. The base weight for teacher sampling is generated by taking the base weight for school sampling (the reciprocal of the probability of selection of the school), adjusting for sampled schools for which a TLF is not obtained, and multiplying this by the reciprocal of the probability of selection of the teacher within the school (from the TLF). Next, a nonresponse adjustment factor for nonresponse is calculated and applied based on a weighting cell adjustment. Weighting cells are developed using tree search algorithms. These cells are selected to be homogenous in response propensity within cells and heterogeneous in response propensity across cells. The adjustment is the inverse of the weighted response rate within each cell, and each respondent in the cell receives this adjustment. Nonrespondents are given weights of zero and the respondents are reweighted to represent the nonrespondents. The variables examined for potential bias were the same as those used by the tree search algorithms.

Imputation. Cases with “not answered” values for items were imputed. Two imputation approaches were used in the 2015–16 NTPS and 2017–18 NTPS. First, donor respondent methods, such as hot‑deck imputation, were used. Second, if no suitable donor case could be matched, the few remaining items were imputed using mean or mode from groups of similar cases to impute a value to the item with missing data. After each stage of imputation, computer edits were run again to verify that the imputed data were consistent with the existing questionnaire data. If that was not the case, an imputed value was blanked. In these situations, Census Bureau's analysts looked at the items and tried to determine an appropriate value. Edit and imputation flags, indicating which edit or imputation method was used, were assigned to each relevant survey variable.

Changes Over Time
Some changes have been implemented since the first 2015–16 cycle.

• In the 2017–18 cycle, private schools, principals, and teachers were added into the target population.
• In the 2017–18 cycle, both national and state-level estimates were supported for public schools, principals, and teachers, while the 2015–16 cycle produced nationally representative estimates only.
• The 2017–18 NTPS collected and reported information on the National School Lunch Program differently than previous years of both NTPS and SASS. Rather than asking for a count of K-12 students approved for the program, schools were asked to report the percentage of K–12 students approved for the program. Additionally, schools that did not participate in the program used to be treated as a separate category for reporting purposes, but in this cycle, are grouped with schools that participated in the program but had no students approved for the program.

Future Plans
NCES plans for the NTPS to be collected on a two- or three-year cycle. The next planned collection phase is 2020–21.