Ever wondered how an opinion poll works? Jack Combs, researcher at the Taubman Center since 1985, tells us how the numbers are gathered and crunched.
1 Questions are formulated by researchers at the Taubman Center. “Ninety percent of the questions we ask are standard questions asked by other polling organizations,” says Combs. Questions are standardized because the phrasing of questions can strongly affect responses. Questions must avoid “leading” the respondent to a particular answer. Survey topics address topics of the day, which lately has been the economy. Every survey tracks the performance of the president as well as congressional members and key local politicians.
2 A random sample of Rhode Island voters is drawn. Voter lists are provided by the Secretary of State’s office. In Rhode Island, 88% of eligible voters are registered to vote, providing a healthy pool to choose from. The sample is drawn proportionately from Rhode Island’s 39 cities and towns. Providence, with 13.7 percent of registered voters, would be 13.7 percent of the sample. Little Compton, with .4 percent, would provide .4 percent of the sample, and so on.
3 Student workers are hired and trained. Interviewers must read from a script to ensure that questions are asked in a consistent manner, thereby ensuring accurate results.
4 Survey is conducted. Surveyors receive their call lists -- a page full of phone numbers -- with no other information attached. Surveyors manually dial the numbers and get an answer about every fifth call. Respondents must be over age eighteen and complete the entire survey, which typically takes ten minutes or more. The survey sample is complete when the number of completed surveys reaches five hundred to six hundred. While this number may seem small, statistical research has proven that for a state Rhode Island’s size, five hundred or more responses will yield a margin of error no greater than 4.5 percent. (To understand how a sample size works, see the National Statistical Service website sample size calculator page).
5 Data is analyzed using statistical package for social sciences (SPSS) software. For every Taubman poll, surveyors collect four variables — party affiliation, gender, age and educational attainment. For some polls, other data (such as race or ethnicity) may be recorded as well. These variables allow researchers to cross-tabulate data to find out how demographic factors affect responses. For instance, in the Taubman Center’s February 2012 poll about the revised contraception policy, age and gender were significant factors in how people responded.
6 Poll results are released to the public via the media. The news media usually focuses on the results on timely topics as well as any swings in approval ratings for elected officials.