Regression Poll February 2023

This page first posted 9 February 2023

Pollster Find Out Now and election experts Electoral Calculus have run a MRP poll on voting intention for Westminster. This was a large-scale poll, involving over 28,000 respondents, carried out from 27 January – 5 February.

The poll asked GB residents whether and how they intend to vote if there were an imminent general election. It also asked voters who would make the best Prime Minister and their attitudes to Conservative sleaze

The headline voting intention is shown in the highlighted column:
PartyVote share at GE 2019Previous poll Feb 2022Previous poll Sep 2022Current Estimated Vote Share (pc)Estimated Change (pc)

Footnote: the predicted vote share percentages differ from the basic voting intention from the poll, as the regression techniques and models operate differently from classic polling analysis. This modifies the parties' vote shares slightly. See below for analysis of don't knows and refuseds.

This gives an estimated Labour lead over the Conservatives of 25pc. That is very significant lead, which would lead to a Labour landslide in a general election.

The regression techniques of MRP polling allow predictions to be made for each individual seat in GB. The number of seats predicted to be won by each party is shown in the next table.

PartyNumber of Seats
at GE 2019
Number of Seats

The prediction is that the Conservative would face electoral near-annihilation, and would not even be the official opposition. SNP Westminster leader Stephen Flynn would be the leader of the opposition.

Conservative losses worse in stronger seats

One reason that the Conservatives are doing so badly, is that the poll shows that Conservative losses are worse in stronger seats.

To show this, we looked at voting intention of those voters who lived in either very strong or quite strong Conservative seats (see Technical Details below for these definitions). Because the overall poll was so large, there were around 5,000 respondents in each of these smaller groups.

Con vote share at GE 2019Current Con vote share from pollAdditive changeMultiplicative change
All GB seats45%23%−22%−49%
Quite Strong Conservative seats57%30%−27%−47%
Very Strong Conservative seats64%34%−30%−47%

Across the whole country, the Conservatives are predicted to lose 22% support, as they decline from 45% at the last general election to 23% today. But in quite strong seats, they lose 27% which is more than the national average. And in very strong seats they lose 30% support, which is even bigger. The Conservative losses are worse in their stronger seats, which helps explain why they are predicted to lose so many seats.

Another way of seeing this is to look at the multiplicative change, which is the fraction of former Conservative voters who are deserting the party. (For example, nationally that fraction is 22%/45% which is 49%). In all three cases, this is around 47% to 49%. In other words, nearly half of all former Conservative voters have withdrawn their support. And this is as true in the strong seats as it is nationally.

Additive and multiplicative changes to CON vote share by seat type

That could be even worse for the Conservatives than 1997. Even in Tony Blair's Labour landslide, Conservative support held up better in their stronger seats.

Best Prime Minister

We also asked people "Who do you think would make the best Prime Minister?". Responses are shown in the table below

ResponseAll votersCON
voters in 2019
voters in 2019
Keir Starmer34%12%67%
Rishi Sunak18%44%3%
Don't know48%45%3%

Table: Excludes those who refused to say

Overall, the most popular option was "don't know" on 48%, with Keir Starmer in second place on 34% and Rishi Sunak behind on 18%. Among those who voted Conservative in 2019, about as many people said Rishi Sunak as said don't know.

This underscores the voting intention figures which show that around half of former Conservative voters are disillusioned with the party. It also suggests that Keir Starmer does not enjoy huge personal popularity, despite Labour's position in the polls.

Conservative Sleaze

We also asked people "Do you agree or disagree that the Conservatives are fundamentally corrupt?". The responses are given in the next table.

ResponseAll votersCON
voters in 2019
voters in 2019
Somewhat agree/
Strongly agree
Somewhat disagree/
Strongly disagree
Neutral or don't know24%32%7%

Table: Excludes those who refused to say

A clear majority of all voters agreed at least somewhat with the proposition. Even among Conservative voters opinion was split in three directions, with some agreeing, some disagreeing and some uncertain. Labour voters were much more certain, with 91% agreeing.

Predicted seat changes by MP

Some notable MPs are in danger of losing their seats, including the Prime Minister and fifteen other cabinet ministers:

Also Boris Johnson (former Prime Minister), Liz Truss (former Prime Minister), and Kwasi Kwarteng (former Chancellor).


Chris Holbrook, CEO of Find Out Now: "Our prediction makes shocking reading for Conservatives, who may yet take some comfort in the lack of personal support for Keir Starmer, and a hope that the shift can be undone before polling day. Mending perceptions of corruption may be their best hope."

Martin Baxter, CEO of Electoral Calculus: "The Conservatives have been far behind in the polls for the last four months, with little sign of improvement. They have lost support across the country, particularly in traditionally strong Conservative areas, which bodes very badly for the next general election." "That election could be a near-wipeout, and worse than 1997, with the Conservatives not even being the main opposition party."

Technical Details

Find Out Now polled 28,191 GB adults online between 27 January – 5 February 2023. The sample was weighted to be representative by gender, age, social grade, other demographics and past voting patterns. Regression techniques were used to infer projected seat results.

The "Very strong" Conservative seats were defined as the top 120 seats by Conservative vote share at the last general election. The sample size of this sub-group was 5,064. The "Quite strong" Conservative seats were defined as the next 120 seats ranked by Conservative vote share at the last general election. The sample size of this sub-group was 4,988. Find Out Now and Electoral Calculus are both members of the British Polling Council and abide by its rules.

Full Data Tables are available for download as an Excel spreadsheet.

Treatment of Don't Knows and Refuseds

The treatment of respondents who didn't clear answer the voting intention question is getting more important, as many respondents now do this. This particularly effects Conservative voters, so the correct treatment can influence the headline voting intention figures.

Note that this does not affect those respondents who have indicated that they probably will note vote. The focus of this analysis is on those who do not know or prefer not to say how they will vote or not.

Traditional polling for voting intention works in two stages, given a set of respondents with valid demographic data.

The headline voting intention is then based on the filtered respondents and their weights.

We call this method "early quota", because the quota weights are calculated early in the process, before non-definite respondents are removed.

But we could run this process in the other order:

We call this method "late quota" because the quota weights are calculated late in the process.

The difference between these two methods could be important if there are a large number of don't knows or refuseds. At the moment we have 16% of all voters, and 26% of Conservative voters, who don't have a clear VI.

The "early quota" method effectively assumes that DKs/refuseds will not vote. The "late quota" method effectively assumes that these voters will vote in the same way as people with similar characteristics who did express an opinion

For information, alternative tables have been produced for this poll showing the effect of the "late quota" method. It produces less extreme leads for the Labour party, showing a 25% lead rather than a 28% lead. It is also more consistent with the MRP method, which has similar assumptions about non-respondents.

Electoral Calculus

Electoral Calculus is a political consultancy specialising in quantitative analysis and modelling for electoral and other market research projects. Its pre-poll prediction for the 2019 general election was the most accurate published forecast. It was founded by Martin Baxter, its CEO.

Electoral Calculus is a member of the British Polling Council.

Find Out Now

Find Out Now gathers poll responses from Pick My Postcode, a daily panel from 2.8 million members. Highly profiled respondents can be targeted instantly, and at scale to deliver reliable results fast.

More than 124 million responses have been received to Find Out Now's polls since it launched in November 2018. Find Out Now are Market Research Society Company Partners and a member of the British Polling Council.

Regression Polling

Modern polling analysis often uses statistical regression techniques to get more accurate and geographically detailed results. Also called MRP (multi-level regression and post stratification) they have been used successfully by Electoral Calculus and other pollsters to predict general elections, local elections and the 2019 European elections.

These techniques work by spotting patterns between people's demographic characteristics and their likelihood to vote for various parties.