Teachers versus the Public: What Americans Think about Schools and How to Fix Them

Appendix A: Data Collection and Analysis

The findings reported in this volume are based on seven annual surveys of nationally representative samples of the U.S. adult population conducted from 2007 to 2013. The surveys were conducted under the auspices of Harvard University’s Program on Education Policy and Governance (PEPG) and Education Next: A Journal of Opinion and Research. The surveys were administered by the polling firm Knowledge Networks (KN). KN maintains a nationally representative panel of adult survey respondents, obtained though list-assisted, random-digit-dialing sampling techniques, who agree to participate in a limited number of online surveys. Although surveys are administered online, the sample is not limited to current computer owners or users with Internet access because KN offers panel members without access free Internet access and a WebTV device that connects to a telephone and television. When recruiting for the panel, KN sends out an advance mailing and follows up with at least fifteen dialing attempts. The panel is updated quarterly. Detailed information about the maintenance of the KN panel, the protocols used to administer surveys, and the comparability of online and telephone surveys is available at www.knowledgenetworks.com/quality/. Participants in the Education Next–PEPG surveys are randomly sampled from KN’s panel of respondents . Several of the surveys included oversamples of various subgroups (for example, parents and teachers). For each survey, post-stratification population weights are calculated in order to adjust for survey nonresponse as well as for subgroup oversampling. Those weights ensure that the observed demographic characteristics of the final sample match the known characteristics of the national adult population. All analyses presented in AP P EN DIX A Data Collection and Analysis 125 the text are based on weighted data. Reported percentages do not always add to 100 as a result of rounding to the nearest percentage point. Because respondents’ geographic locations were known prior to the time that the survey was administered, it was possible to link respondents to personalized information about local schools either for use in the survey or for analysis after data were collected. So, for example, we obtained information on the amount of per-pupil spending in respondents’ local school districts. In some cases that information was used to compare perceptions of spending to actual spending; in other cases, the information was provided to respondents as they completed the questionnaire. In order to secure accurate matches between respondents and information about local districts, we matched survey respondents to school districts using either census blocks or zip codes. When we relied on zip codes, we could not match some respondents to a unique school district. For such respondents we calculated the average per-pupil spending levels for each district that served the relevant zip code, weighted by the districts’ population size. For the experiments involving teacher salaries, only state-level data were available. We were able to match all survey respondents to the states in which they resided. Data on per-pupil spending came from the Local Education Agency Finance Survey of the National Center for Education Statistics ’ Common Core of Data. Data on teacher salaries came from the American Federation of Teachers publication “Survey and Analysis of Teacher Salary Trends.” In each survey, the most recent financial data available were used. The survey administered in 2011 included many more questions, carried out more experiments, and oversampled many more groups than previous iterations of the survey, allowing for a more extended examination of many issues than in previous years. As a result, the data reported in this volume lean heavily on the 2011 results, even though it makes use of data from other years when doing so illuminates a topic. If results come from a year other than 2011, that fact is specifically indicated in the text; otherwise the reported numbers are from the 2011 survey. —2007. The 2007 survey was conducted between February 16 and March 15, 2007. Data were collected from a nationally representative stratified sample of 2,000 adults (ages 18 years and older). The sample consists of 1,482 non-Hispanic whites, 233 non-Hispanic blacks, and 126 Appendix A 171 Hispanics. The survey oversampled parents of school-age children, who constitute 811 members of the total sample. —2008. The 2008 survey was conducted between February 16 and March 15, 2008. The sample consists of 2,546 non-Hispanic whites, 250 non-Hispanic blacks, and 239 Hispanics. It includes an oversample of approximately 700 public school...