Election and policy polling
Google Surveys is frequently used by pollsters to conduct internet-based polls. To help you produce quality polls, we've provided the following information and suggested guidelines.In this article:
Election polling conducted by professional pollsters is often more sophisticated than simply asking the population who they are going to vote for in the next election. Pollsters employ a variety of techniques to reduce bias introduced in the sampling methodologies or the designs of their surveys.
Pollsters frequently use voter files, lists of all the registered voters in given areas, as the sample frames for their research. Doing so lets them only poll likely voters, giving them a more accurate view of the potential outcomes.
Pollsters also employ firms that conduct phone-based polling. This mode of surveying the population lets the person asking the question push a user to answer one way or another if they are unsure by asking them who they are leaning toward. Sometimes this technique is useful to ensure that the proportion of “I don’t know” or “Unsure answers” is reduced and the intentions of voters on the fence are better understood.
Pollsters typically weight the data in proprietary ways to reduce the sampling error. This is done so that the sample better matches the population, taking into account socioeconomic variables such as age, race, education levels, and income.
Because Google Surveys does not employ many of these techniques and only weights the data based on age, gender, and location, we’ve used survey design to approximate the techniques employed by professional pollsters. Because our reach of internet users is so large, the sampling methodology we use doesn’t need many of these tricks to provide accurate results, as proven in the 2012 United States Presidential election.
We’ve found that using the survey-questionnaire design detailed below, we were able to accurately find likely voters and measure their voting intentions.
- Targeting: General Population
- Question 1 screens in likely voters:
- Type: Single Select
- Question Text: “How likely are you to vote in this year's Presidential election in November?”
- Answer Options (displayed in order):
- “100% likely" (screen in)
- “Extremely likely” (screen in)
- “Somewhat likely”
- “Not very likely”
- "I prefer not to say"
- Question 2 reveals voting behavior:
- Type: Single Select with other
- Question Text: “Suppose the presidential election were held today. Who would you be more likely to vote for? If unsure, who are you leaning toward?
- Answer Options (displayed in a random order):
- Donald Trump, the Republican
- Hillary Clinton, the Democrat
- Gary Johnson, the Libertarian
- Still undecided on who to vote for
- Other : (free text write-in)
Because of the way we ask these questions (self-administered surveys on the web without the help of a human interviewer), we tend to see a higher proportion of “Unsure” and “Other” answers than one might get with telephone or face to face interviews.
Many of the same principles apply to policy polling on Google Surveys. Generally, researchers want to either find out what the public thinks in order to support or refute a political party’s position, or they want to understand what voters think of an issue in order to support or refute a candidate’s position or to gauge how the candidate’s own party thinks about a particular issue.
When surveying the public, it is often useful to ask for a user’s political affiliation and screen in respondents that meet certain criteria. When surveying voters, we suggest you use the voting-likelihood question detailed above.
- Question 1 segments by political affiliation:
- Type: Single Select Question
- Text: “In politics TODAY, do you consider yourself a…?
- Answer Options:
- I prefer not to say
- Question 2
- Question Text: “How do you feel about the recent federal tax reform for individuals?”
- Answer Options (randomly reverse order):
- Strongly Favor
- Somewhat Favor
- Neither Favor nor Oppose
- Strongly Oppose