Election Polling Season: A Comparative Analysis of Sampling Methods in the UK

Traditional Sampling Methods, Enhanced Sampling Techniques and Methodological Comparisons

As election fever grips the UK but also in whole Europe, with numerous countries heading to the polls, the demand for accurate and reliable polling data is at an all-time high.   

The Boundary Commissions for England, Scotland, Wales, and Northern Ireland have recently reviewed the UK’s constituency boundaries to ensure fair representation by making constituencies roughly equal in size and preserving local connections. These revisions took into account population fluctuations and changes in electoral ward boundaries. Their final recommendations were published in June 2023, and the updated boundaries will be implemented in the 2024 UK general election.

With approximately 90% of constituencies undergoing various modifications, it’s crucial to consider effective sampling methods for UK election polls. In this context, we explore 3 prominent options: the traditional listed sample, the traditional Random Digit Dialing (RDD) method and the enhanced RDD sample.

The United Kingdom comprises 650 parliamentary constituencies, each represented by a Member of Parliament (MP) in the House of Commons. Recent amendments to constituency boundaries have seen some wards transferred between constituencies, while other changes have been more extensive, including minor boundary adjustments and even the dissolution of constituencies, leading to the creation of new ones.

The electorate quota is the average number of voters each constituency should have if evenly distributed across the UK. Variations in voter numbers are due to population changes and more flexible rules that allow for different quotas and greater discretion to account for local geography or community ties.

Listed sample v.s. RDD sample

Listed samples, commonly employed in election polling, typically focus on reaching eligible and registered voters through predefined lists. However, these lists may overlook individuals not captured within their parameters or contain outdated information, leading to potential inaccuracies in the sampling process. In contrast, RDD (Random Digit Dialing) samples for election polling offer a more impartial approach, as every phone number within the sampled area has an equal chance of being selected. This method significantly reduces bias arising from unclear sources and ensures a more comprehensive coverage of the population. Moreover, RDD samples are particularly valuable for capturing underrepresented demographics, such as the younger population aged 18-29, who may not be adequately reached through traditional sampling methods. By incorporating RDD sampling, election polls can enhance their accuracy and inclusivity, better reflecting the diverse electorate.

Random Digit Dialing (RDD) is also a common method used to select survey participants by randomly generating phone numbers and contacting individuals for surveys or polls. While RDD aims for unbiased representation, combining it with social media data can boost representativeness.
Mobile RDD facilitates wider outreach, especially among younger voters who are more likely to use mobile devices as their primary communication tool. On the other hand, RDD landlines are more suitable for reaching older demographics or rural populations where landline telephones are still prevalent. The choice between the two methods impacts the demographic composition and targeting level of the sample. Mobile RDD tends to capture a more diverse and younger demographic, while RDD landline may skew towards older and more rural populations. Consequently, understanding the demographic characteristics and communication preferences of the target population is crucial in selecting the appropriate sampling method for election polls.

Big Data enriched RDD sample

A ‘Big Data enriched RDD sample’ merges RDD with insights from social media platforms. This method supplements RDD’s demographic data with additional information from social media profiles, providing a more nuanced view of the target population. By tapping into social media, researchers gain insights into demographic subgroups, including hard-to-reach younger audiences aged 18-29. This approach improves the sample’s accuracy and ensures a more comprehensive understanding of the population.

While RDD sampling stratified at the constituency level faces significant challenges in the current electoral landscape of the UK. Using RDD samples stratified at the constituency level is not the most effective method for sampling the UK for election polls. Constituency boundaries are composed of wards, which are administrative divisions within cities or boroughs. Constituency boundaries often cut across wards, creating complexities in accurately stratifying samples. This can lead to underrepresentation or overrepresentation of certain areas within a constituency. Wards, being administrative divisions, do not align neatly with the UK numbering plan and the demographic and social characteristics of the electorate, leading to potential biases in sampling.

Quota v.s. Probability-Based

Quota sampling and probability-based sampling are two distinct methodologies used in election polling in the UK. Quota sampling involves selecting participants to match specific demographic characteristics, ensuring the sample reflects the population’s diversity. This method is quicker and less expensive but can introduce biases as it relies on the judgement of the interviewer. In contrast, probability-based sampling ensures that every individual in the population has a known and non-zero chance of being selected, resulting in a more statistically robust and unbiased sample. While more time-consuming and costly, probability-based sampling provides more reliable results, making it the preferred method for accurate election polls in the UK.

Exploring alternative approaches such as using an enhanced RDD and listed sample can help pollsters achieve more accurate and representative samples. By leveraging B2C data, which can be enriched with demographic and geographic information, pollsters can create a more precise and targeted sampling framework. This approach can overcome the limitations of traditional RDD sampling by ensuring that the sample better reflects the actual distribution and characteristics of the electorate within each constituency.