Consumer Insights Tracker Technical Report: Sampling

Sampling approach

All respondents who take part in the survey are drawn from the YouGov panel of over 400,000 active panel members who live in the UK. YouGov maintains an engaged panel of respondents who have specifically opted-in to participate in online research activities. This includes actively recruiting hard-to-reach respondents, such as younger people and those from ethnic minorities, via a network of partners with access to a wide range of online sources that cater to these groups. These partners have specific experience in recruiting these audiences for online activities. The sources include search engine optimization (SEO), affiliate networks, niche websites, and growth hacking techniques such as panellist refer-a-friend campaigns and social networks. All recruitment sources, for hard-to-reach audiences or anyone else, are monitored to ensure that respondents are always profiled, responsive and engaged in the survey experience.

The YouGov panel is large enough to select nationally representative samples that reflect the actual breakdown of the population on the key demographics of age, gender, region, social grade and education level.

At the start of the project a nationally representative sample of people in England, Wales and Northern Ireland was constructed. To qualify for the survey respondents needed to meet two criteria:

• Be aged 16+ years of age
• Live in England, Wales or Northern Ireland  ^{(footnote 1)}

Each month, eligible panellists (see further details below on sampling) are sent an email inviting them to take part. Only panellists who are invited can participate in the survey, and individuals can only respond once each time they are invited. Participants are given a brief introduction to the survey before responding. They are not made aware that the survey is being conducted on behalf of the FSA.

The sampling approach used for the Consumer Insights Tracker is quota sampling, with quotas taken from the 2021 Census. Quota sampling is a non-probability sampling method that involves dividing the population into sub-groups. The relative proportion of each sub-group as a percentage of the total sample is based on known characteristics of the overall population. Responses are collected until the ‘quota’ for each sub-group is reached. This is a standard sampling approach used in online panel surveys to capture a representative population. The sample for the Consumer Insights Tracker is structured to be representative of the England, Wales and Northern Ireland population by the following variables:

• Age
• Gender
• Social grade
• Region
• Education level

A sample boost is also applied in Northern Ireland (to achieve 100 respondents), to improve the representation of this group and to enable rudimentary demographic comparisons between countries. An exclusion criterion is also applied to the sample, which prevents any respondents from participating in the survey if they have done so in any of the six previous waves.  

Sample size and weighting

The overall sample size for each wave is approximately 2,000. This sample size is sufficient to ensure that robust analysis can be conducted across various demographic sub-groups. It also ensures that margins of error within survey results are minimised, enabling statistically significant variations to be identified across and within waves.

Weights are applied to the data to make it more representative of the population of England, Wales and Northern Ireland.  Weighting involves adjusting the contribution of individual respondents to ensure it is given the correct relative influence based on the demographics of the target population. Data from the 2021 UK Census from the Office for National Statistics is used to do this, with weighting applied according to gender, age, education level, region, and social grade. Weighting on these variables is common practice across similar surveys of this nature, as it ensures that the data is representative of the target population, while also maximising weighting efficiency.

Random Iterative Method (RIM) weighting is used. RIM is used when there are a number of different standard weights that all need to be applied together. This weighting method calculates weights for each individual respondent from the targets and achieved sample sizes for all the quota variables. RIM weighting is an iterative process, whereby it recalculates the weights several times until the required degree of accuracy is reached. All weights are capped at six, and a weighting report is produced for each wave. The advantage of using a RIM weighting approach is that the weighting can include a greater number of variables, and it is not necessary to have targets for all the interlaced cells.

Weighting is applied at the end of the data processing phase on cleaned data (see the Quality Assurance section for further details on data cleaning).  

Table 2 provides an example of how data is weighted from the March 2024 wave of the survey. The unweighted base shows the number of completed surveys and the weighted base shows the adjustments that have been made to correct for any sample bias.

Table 2:  Demographic variables and their weightings