Report page 153

Principals’ survey responses have been weighted to be representative of the national population of public school principals at the time the survey was administered.

To provide a weighted sample similar to the population of K–12 public school principals in the United States, we created sampling weights.

The weighting process accounts for the probability of selection into the survey from the panel and the likelihood an invited principal completes the survey (the response weight), and these likelihoods are combined and calibrated to reproduce the population distribution of K–12 educators.

More specifically, the final analysis weights used in our analyses are the product of the following three interim weights:

calibrated weight of the sampling frame (a calibration weight that assigns a weight for each panelist based on individual- and school-level characteristics so that the sum of the weights along the calibration factors closely match the national characteristics of the national population of public school principals based on the NCES’s National Teacher and Principal Survey)

sample selection weight (the inverse of the probability of selection into the survey sample using the panel as the frame)

survey response weight (the inverse of modeled probability of a principal completing the survey. The response adjustment is important to eliminate observed sources of response bias).

The products of these weights were subsequently recalibrated and trimmed if necessary.

As shown in Table B.3, both before and after weighting, survey respondents look similar to the population of U.S. public school principals on a variety of person-level and school-level characteristics.

We report samplewide and subgroup-specific means and proportions of variables of interest, weighted using the survey weights we describe in the previous section.

Because principals’ perceptions and experiences can vary depending on their school context, we examined differences in educators’ responses by school characteristics (in addition to their state policy context).

More specifically, we examined differences by school grade level, school and district enrollment size, locale, poverty status, and student racial/ethnic composition.

We obtained data on school and district demographics by linking survey data files to the Common Core of Data (CCD) and the Education Demographic and Geographic Estimates (EDGE) School Neighborhood Poverty Estimates, both issued by NCES.

school grade level (elementary, middle, and high)

school locale (urban, suburban, and rural)

school student racial/ethnic composition (schools in which more than half of students were Black, Hispanic, Asian, Pacific Islander, American Indian/Alaska Native, or of two or more races were categorized as having majority students of color, with the remaining schools categorized as having majority White students.)

Our definition of locale aligns with the four-category definition of locale used by NCES, with the exception that we collapsed those located in towns into the rural category.