Opinion polls gauge and compare a specific group’s views and opinions on a topic. Opinion polls can be delivered in several forms, including questionnaires, face-to-face interviews, telephone surveys and online surveys.
Opinion polling methods utilise questions that are developed prior to carrying out the poll, and have no scope for deviating from them. Participants in the poll do not have the opportunity to provide substantial opinions or attitudes on areas outside of those in the poll questions.
Click here for details of the tool characteristics.
Best Practice Examples
The Department of Energy and Climate Change undertake public attitudes tracking omnibus surveys each quarter to understand and monitor public attitudes on their main business priorities, helping them to address public concerns in a fast-changing energy sector.
When to use and not use opinion polling
When to use opinion polling
• When you wish to gain an understanding of the opinion of a large population
• When the polling topic is well understood by the public, and requires little clarification of meaning or reflection from the participant
When not to use opinion polling
• When it is not possible to ensure the sample is representative of the target group
• When you wish to understand the underlying values of the public
• When you wish to track the change in respondents’ opinions over time
• When you would like a dialogue between responder and those carrying out the survey
• When the direction of research is iterative or requires steering
How to conduct an opinion poll
1 Articulate what can and should be measured
Opinion polls are structured so that responders provide discrete responses to set questions, after which the discourse between the responder and policy maker ends. A thorough understanding of the public’s attitudes and values is unlikely to be gained from opinion polling, for several reasons.
Polls do not offer the responder the chance to elaborate on their interpretation of the question, nor how they have arrived at their response. Responders may lack information on the background or context of the polling topic or time to reflect on their answers.
Opinion polls are therefore more suited to measuring behaviour, opinion, knowledge and demographic characteristics , for which questions are easier to devise and are less prone to debate.
2 Design the wording and format of questions
Opinion poll questions should be phrased such that they can be understood and measure the opinions they are intended to. For questions with clear answers such as age and educational qualification, this is a relatively straightforward task.
Other questions should choose wording that avoids bias or misleading the responder. Bias may occur when one side of a topic is presented more favourably or emotive language is used.
The responder must have a clear understanding of what you would like their opinion on. If a question is ambiguous, complex or contains technical jargon, this is less achievable. A question that contains multiple propositions within it is similarly confusing, as responders may be unsure which they are responding to.
Design and formatting factors to be aware of include question ordering, where preceding questions may frame the answer to a given question. In categorised questions you should also ensure categories cover all responses, and that scaled questions are sensitive enough to reflect the range of different responses. A scale of 1 to 3 is less likely to capture the variance in a response than a 1 to 10 scale for example.
3 Determine what a representative sample looks like, and how you might achieve it
To be able to generalise the results of a poll to the population whose opinions you would like to understand, it is essential that the sample used is representative of this population.
Random sampling is where everyone within a target population has an equal chance of being selected for polling.
Cluster sampling can be used for scenarios where random sampling is impractical or disproportionately costly. Often used for face-face polling, cluster sampling uses representative samples in a particular neighbourhood of the population wanted to be studied. This avoids the potential costs of travelling large distances to conduct a small number of interviews to make a truly random sample of a large area.
Quota sampling involves setting quotas based on demographics, such as race or gender, and seeking out people who as a group match those characteristics and meet the set quotas. YouGov uses quota sampling for its British Election Study. In doing so their sample is weighted based on factors such as age, social grade and even newspaper read to ensure the polling sample is representative of the voting population.
It is also worth noting that the effects of quota sampling can be achieved using random sampling and demographic questions that can be used to weight data afterwards. If it is costlier to find a non-random sample then to apply weightings afterwards, this may well be the better option.
4 Implement the opinion poll
If the poll consists of a small amount of questions, the cheapest option is to buy some questions on an existing survey, or outsource to a polling provider such as YouGov’s omnibus service.
If the survey is larger scale, carrying out independently will give more control of the quality, although costs will likely be higher. Online, postal and email surveys are naturally suited to larger polls as they do not require individual staff to conduct each poll, and therefore involve less time and man-hours per participant. Such surveys are however more susceptible to lower overall response rates and responder bias. There is no guarantee that people who have the time and interest to respond to such poll formats are representative of the population as a whole .
If the topic of the poll is socially sensitive or intrusive, techniques such as face-to-face interview or telephone poll that involve direct interaction between interviewer and interviewee are less suitable, as responders may alter their responses based on social cues and biases.
5 Consider how to use the data
For closed questions, a range of software can quickly analyse poll responses and identify trends and topline results.
• Opinion polling gives people who do not usually have access to the media an opportunity to be heard
• The scale of some opinion polls means they can generate statistically significant data for sub-groups of the public
• Standardised surveys can be delivered in many formats in a cost-effective way
To be aware of
• Caution should be advised when polling complex issues with which participants are unlikely to be familiar. Results may be misleading if participants misinterpret questions or select an answer that only roughly corresponds to their actual views
• Closed-ended survey questions may not capture the full range of participant views
• Opinion poll data is susceptible to bias resulting from factors such as the wording and order of questions, and such effects should be considered carefully