November 7, 2012

No Consensus View on Election Outcome

Pew Research Flash Election Reaction Monitor

More about the Google Consumer Survey Method

The Google Consumer Surveys method has the ability to capture reactions from a broadly representative, though non-probability, sample of internet users in a relatively short period of time. Internet users are sampled by selecting a stratified sample of people visiting the websites of a diverse group of more than 80 publishers that allow Google to ask one or two questions of visitors to their site. This sampling procedure is different from other internet surveys that rely on people who “opt-in” to participate or that randomly survey respondents who have agreed to be part of a pre-recruited online panel.

The data are weighted to match national parameters for internet users on age, gender and region or state; these demographic characteristics are inferred based on the types of websites the users visit as recorded in their DoubleClick advertising cookie and their computer’s internet address. Because the final sample is not a probability sample of all internet users, it cannot be assigned a margin of sampling error. The response rate for a political question typically ranges from 15% to 30%.

One restriction of this method is that only two questions can be asked of any individual respondent, limiting the ability to analyze the results of multiple questions among the same group of respondents. There also are fewer demographic variables (inferred age, gender, location and income) available for analysis than from traditional survey questionnaires. One advantage of Google Consumer Surveys is that questions can be fielded very quickly – within a matter of hours or one day – making it possible to monitor immediate reaction to breaking news events.

Comparative tests on more than 40 political, social and technology use questions over the past few months have found that the Google method produces results that are quite similar to the results of Pew Research Center’s standard RDD dual frame telephone surveys. The median difference across questions tested was 3 percentage points and the mean difference was 6 percentage points. In addition, assessments of the composition of Google datasets also suggest that they closely match the U.S. internet population demographically. A report on the comparative testing, along with a more extensive description of the Google methodology, is available here.

Question wording and full results are available in the topline and on the Google Consumer Surveys site for one-word reactions among Obama voters and Romney voters, and for the overall reaction to the outcome.

For more on the Pew Research Center’s work with Google, see this Q&A with Scott Keeter, Director of Survey Research.