In line with our obligations to the British Polling Council and the Market Research Society, YouGov are publishing the data, methodology, and question wording for the two MRP polls ran by YouGov, commissioned by the Conservative Britain Alliance, with David Frost named as the contact, and published on Wednesday 24 January 2024.
YouGov interviewed 14,110 adults across the country between 22 December 2023 and 4 January 2024 and asked them to complete the following hypothetical thought experiment.
First, respondents were asked a standard ‘best prime minister’ question.
Which of the following do you think would make the best Prime Minister?
- Rishi Sunak
- Keir Starmer
- Not sure
Respondents were then later presented with the following hypothetical scenario:
Please imagine that we are well into 2024, and the next UK general election is being held. Suppose by this time, the Conservatives had replaced Rishi Sunak and elected a new leader and prime minister, and they had made the following changes:
- Taking a much tougher approach on people coming to the UK illegally on small boats from France
- Introducing tough new measures on crime and anti-social behaviour
- New measures which significantly reduced legal migration
- Bringing in new tax cuts for working people
- Successfully getting waiting times for operations with the NHS falling
Respondents were then asked the following question:
And in this hypothetical election, which of the following would you prefer to be prime minister?
- The new Conservative leader
- Keir Starmer
- Not sure
The experiment does not ask (nor suggest) anything about vote intention or what would likely happen at an election given this entirely hypothetical scenario. It only asks which of two options respondents think would make the best prime minister.
The results of both ‘best prime minister’ questions were then projected down to the constituency level in England and Wales using Multilevel Regression and Post-stratification (MRP for short). MRP constituency models first estimate the relationship between a wide variety of characteristics about prospective voters and their opinions – in this case, ‘which of the following do you think would make the best prime minister' – in a ‘multilevel model’. It then uses data at the constituency level to predict the outcomes of seats based on the concentration of various different types of voters who live there, according to what the multilevel model says about their probability of voting for various parties (‘post-stratification’). In this instance, 500 draws from the posterior distribution of the multilevel model were used to predict the constituency probabilities. The multilevel model was estimated through 12,000 iterations.
The constituency-level data for the original ‘best prime minister’ question and then the following ‘hypothetical best prime minister question’ can be found here and here.
Hypothetical polling of this nature is complicated and must be reported and interpreted with many caveats. Perhaps most importantly in this particular case, the new hypothetical leader is not a specific person. This means that respondents will project onto them whatever personality, characteristics, leadership skills and qualities, and other policy platforms that they may so choose. Often these images of hypothetical leaders will be contradictory and inconsistent between respondents. It does not tell us much about what would actually happen if a specific person became a fresh leader of the Conservative party. That would require polling, and MRP modelling, of its own.
