MethodologySkip to content Skip to main navigation
The Public Matters
Results are based on online survey conducted August 28 – September 6, 2017 among a national sample of 3,117 adults 18 year of age or older using the Qualtrics Panel. Qualtrics, a marketing research firm, partners with a variety of online panel providers to supply a nationally representative sample. The sample is compiled using overall demographic quotas based on census percentages for representation (i.e., age, gender, race/ethnicity, household income, and census region). To allow greater power for analysis, we over-sampled people who identify as Black, Asian and Pacific Islander, and/or Latinx. The sample is weighted to represent the U.S. adult population living in households or group quarters.
The survey included several quality assurance measures, including attention checks and a speed check. Attention checks asked respondents to mark a specific answer. Respondents who failed one or more of these checks were removed from the final sample.
Most of the survey items were developed by the research team and colleagues at Teachers College. Other survey items were adapted from the NORC General Social Survey (GSS), Gallup and Pew. The survey also included a detailed battery of survey items on sociodemographic characteristics. In addition to close-ended items, the survey asked several open-ended questions, allowing respondents to contextualized and explain their responses in greater detail.
To test for framing/wording effects, the survey contained several experiments. For example, we vary the term used for describing educators as stakeholders in education. We used the term “teacher unions” to situate educators as an organized and political entity. And we used the term “teachers” as a more neutral reference. That is, the experiment included a total of two conditions. Respondents were randomly assigned to each condition.
All surveys are subject to various forms of error. One form is sampling error: the variation in results that is attributable to chance in which members of a population are randomly selected to participate in the survey. For percentages based on the entire sample, the approximate margin of error is +/- 1.8%. For subgroups, the margin of error is larger. For example, the margin of error for Black respondents is approximately +/- 4.5%.
Socio-demographic profile of the sample is available here.