How Prolific accurately predicted the UK general election vote share

Getting an accurate read on how the public thinks and feels is a challenge for many pollsters. But in the days running up to the UK general election, Prolific Polling was able to accurately predict voter intention. 

How did we do it? In this post, we’ll explain how our representative samples and MRP modeling allowed us to accurately predict vote share in a matter of hours.

Polling data that shows how the UK really feels

Every two weeks, Prolific polls a 2,000-person representative sample of the UK population from our pool of 200k+ participants. The sample is representative by age, gender, and ethnicity, based on the latest census figures. We aim to track how the UK public feels about various societal issues, such as AI and the cost of living, and to track voting intention.  

Two days before the UK general election, we increased this sample to 3,000 and asked: “Which party do you intend to vote for in the upcoming general election?” We then ran the responses through our multi-level regression and post-stratification (MRP) model. This allowed us to estimate the national level data by factoring in 2019 general election voting data, making it even more representative. 

The results were highly accurate. Our final vote share predictions closely mirrored the actual outcomes of the UK general election, demonstrating how the Prolific platform can be used to predict the behavior of the UK population. 

How Prolific compared to other BPC members

Prolific is a member of The British Polling Council (BPC), which encourages the highest professional standards in public opinion polling. As BPC members, we’re committed to transparent polling methods, conducted with statistical rigor.   

Out of the 19 members of the BPC, Prolific ranked 6th most accurate for predicting vote share in terms of the average absolute deviation from the observed vote share percentages. We ranked ahead of some big names, including Savanta and Redfield & Wilton.

You can view the results below, which include:

• Rankings of pollsters based on overall accuracy.
• Absolute deviations from the actual observed vote share.

Prolific was the 6th most accurate for predicting vote share

To calculate overall accuracy, we computed the difference between the observed vote share for all major UK parties in the recent general election and the final vote share predictions from each BPC member. We then calculated the average differences between each pollster's observed and predicted vote shares to give us average deviation scores, which we ranked.

Prolific collected final vote share data in 3 hours

While many pollsters use online panels to collect data, none have access to our large, vetted pool of highly engaged participants. 

Prolific’s fast, flexible platform and representative sampling meant that we could get results extremely quickly. We collected our final vote share data in just three hours.    

This approach is also much more cost-effective than traditional polling methods, costing just £450 - and the results speak for themselves. 

Reliable polling for the questions that matter

Getting a reliable read on public opinion can be challenging, especially for timely topics where results are needed fast. However, Prolific’s representative sampling makes it possible - the high accuracy of these poll results and the speed at which they were gathered prove it. 

Voter intention is just the start. With our representative samples and flexible platform, we can accurately track public sentiment on everything from the cost of living crisis to evolutions in AI.   

You can find out more about the methodology behind our polling data and get involved here.