Back to top

Poll Methods and Data

Poll Methods and Data

IEA engaged The Benenson Strategy Group and The Winston Group to design and conduct the national opinion poll. The Benenson Strategy Group provided guidelines for sample composition, including: (1) a base sample of no fewer than 1,000 randomly selected American citizens, to ensure generalizable results (Gelman, 2004), (2) adequate representation of “Education Influencers,” who are known to impact education policy at the local or national levels, specifically parents of school-aged children and Opinion Elites [1] , and (3) oversamples of Black and Hispanic respondents, to ensure poll results represented the perspective of groups traditionally underrepresented in gifted programs. [2]   The recommended minimum sample size for each of these target groups was 150 respondents.

Poll respondents were members of online panel groups associated with companies that collaborate with The Benenson Strategy Group on a regular basis. Individuals who are members of online panel groups “opt-in” to receive invitations to participate in polls through the panel company. The panel company ensures the honesty of panel members and limits the number of polls each panel member completes. Members of panel groups are not informed about poll topics before agreeing to participate, so IEA-P respondents did not know in advance that they would be completing a poll about education issues.

The Benenson Strategy Group screened all potential members of the respondent group. In the process they excluded 493 individuals because they did not meet the screening criteria. Exclusions usually occurred because an individual did not meet the requirements to be part of an oversample. Another 111 respondents began, but did not complete, the poll. [3]   The resulting respondent group consisted of 1414 registered voters, aged 18 and older. Around 22% of the respondent pool reported having gifted children, although some of those children were adults. Demographic characteristics of the full respondent group and targeted subgroups are in Table 1.1; additional demographic data describing poll respondents is available in Appendix B, Table B1.

1. Poll Design 

IEA and the symposium participants worked with The Benenson Strategy Group to design a pilot poll. This pilot poll was comprised of 24 open-ended items and was used to conduct in-depth online conversations using the iModerate™ software with six parents, six educators, and eight Opinion Elites. The pilot group’s answers to these questions determined the language, critical issues, and messaging themes for the more extensive poll. The Benenson Strategy Group, The Winston Group, and additional contributors went through several revisions before the Benenson Group created a final draft which was reviewed by The Winston Group to ensure item validity and political neutrality.

The result was a 76-item poll, which is included in Appendix A. The first 17 items gathered demographic information used to create analysis groups, and the final 11 questions gathered demographic information typically solicited in Benenson Strategy Group polls. The remaining 48 questions comprised the core content of the poll, although several items were divided into several parts. [4] Consequently, the entire sample answered 25 identical questions and another 23 questions with variations in split samples. The distribution of questions across split samples is in Appendix B, Table B2.

Table 1

Demographic Distribution of IEA-P Respondents

    Race/Ethnicity   Race/Ethnicity x Income   Education
            White Hispanic Black      
  Total White Hisp. Black   <$50 $50k+ <$50 $50k+ <$50 $50k+   OE Parents
Weighted n: 1414 1004 156 176   395 580 75 73 96 74   42 424
Unweighted n: 1414 815 281 261   246 554 97 182 128 126   246 690

Note. Hisp. = Hispanic, OE= Opinion Elites. <$50 = annual income under $50,000/year, $50k+ = annual income $50,000 and above. Race/Ethnicity does not include respondents who selected more than one race. Race/Ethnicity x Income does not include respondents who selected “Prefer Not to Indicate.” 

2. Poll Procedure 

The Benenson Strategy Group conducted 1414 25-minute online interviews using the iModerate™ software between December 19, 2016 and January 6, 2017. Where necessary, items were presented in random order to minimize order effects (i.e., when asking for definitions of terms or grades for different aspects of public education) (See Appendix A). Upon completion of the poll, IEA-P respondents received “points,” which could be redeemed at a number of online merchants, consistent with the policies of the online panel companies. [5]

3. Data Analysis 

The aim of this poll was to identify response trends in the American public and to target promising avenues for advocacy. Data were weighted to reflect the demographic make up of the United States, rebalancing the proportions of the oversamples into numerically representative groups. After weighting the data to reflect the demographic makeup of the US, the Benenson Strategy Group prepared a summary based on the weighted data for each question in the poll for the entire respondent group and cross-tabulated by racial/ethnic groups (Black, Hispanic, and White) and Education Influencers (Opinion Elites [6] and Parents).

They also provided standard error of measure at the 95% confidence interval for the full respondent group which is ±2.51%. It is ±6.21% among Opinion Elites, ±3.73% among Parents, ±6.03% among Blacks, ±5.81% among Hispanics, and ±3.33 among Whites. Standard error of measure for split samples, and target subgroups are in Appendix B, Table B3. A set of supplemental data tables is available at:


Gelman, A. (2004). How can a poll of only 1,004 Americans represent 260 million people with only a 3 percent margin of error? Scientific American, 290(3). Retrieved from: of-only-100/

Jefferson, T. (1784). Notes on the state of Virginia. Epilogue: Securing the republic. Retrieved from

Raghunathan, T. E., & Grizzle, J. E. (1995). A split questionnaire poll design. Journal of the American Statistical Association, 90(429), 54–63.

Support for gifted programs vary greatly from state to state. (n.d.). Retrieved from