Multilevel regression and post-stratification (MRP) is a way of producing estimates of opinion and attitudes for small defined geographic areas. It works by combining information from large national samples (for example tens of thousands of respondents) with ONS and census data.

The MR (Multi-level Regression) part

The responses given by respondents are modelled on the basis of their demographic characteristics and what we know about their area (its past voting history, how it voted in the EU referendum, and so on). This is the “multilevel regression” part.

For example, a 23 year old female living in London who works in the media sector and has a university education has a higher probability of being a remain voter than a 72 year old male living in Grimsby who is a retired former fisherman that left school at 16.  

There are elements of a person’s lifestyle, background and life experience that may provide an indication as to their likelihood to vote in a certain way (or choose not to vote at all). “Multi-level regression” examines to what extent each of these elements has an effect on behaviour.

The P (Post-stratification) part

In the subsequent “post-stratification” stage, we use census data to calculate how many people of each demographic type live in each area and combine this with additional relevant contextual information to predict how many of these people will vote for each party (or have a certain opinion).

In this way, the estimates, although they are derived from a national sample, end up being representative of the demographic make-up of each constituency.

What can it be used for?

The MRP technique allows us to model, based on demographic characteristics, an area’s attitude to almost anything (from immigration to consumer behaviours). So the technique will not only assess the likelihood of an individual with a particular set of variables having a specific opinion but also map out how different individuals with different opinions or behaviours are distributed across the country.

Survation recently used MRP during the 2019 General Election to correctly predict a large Conservative majority and call 94.3% of seats correctly. For more information, you can read this blog post by our partner Professor Chris Hanretty here.

An example of Survation’s MRP including our data visualisation tool can be found here. Survation surveyed 20,000 people across the UK for the Conservative Environment Network to access attitudes towards all forms of green energy.

Another example of MRP in practice was the Channel 4 programme “Brexit: What The Nation Really Thinks” which aired on 5th November 2018 and included analysis of a Survation poll with a 20,000 UK sample.

Survation is currently working in partnership with the Department of Politics and International Relations at Royal Holloway, University of London. More information on this partnership can be found here.

More information regarding MRP, it’s varied uses and details of Survation’s quarterly MRP Omnibus can be found here.

If you are interested in commissioning a piece of research which uses the MRP technique or to learn more about Survation’s polling capabilities, please contact Harry Mason on 020 3818 96612 or John Gibb on 020 3818 9661 or email researchteam@survation.com.

Would you like to have your own views represented in Survation’s research work?

Survation’s paid consumer opinion panel is now live.

Simply go to: https://panel.survation.com/ – register (1 minute) and we’ll credit your account with 50 points for simply confirming your email and allocate 100 points for taking a very short survey. Points convert to cash which can be withdrawn each time you reach the equivalent of £50.