Archive for the 'Pollsters/polling' Category


Across all pollsters the Tories are retaining a clear lead

Tuesday, July 7th, 2020

One of the oddest features of current next general election polling is how little there is of it at the moment. The main pollsters of yesteryear, ICM and Populus are not heard of much in this context and even YouGov has only been issuing new polls every few weeks. The last published survey from them finalised its fieldwork on June 12th – which is getting on for a month ago.

Survation, top pollster at GE2017, is producing a survey about once a month while we also have the newcomer, Redfield & Wilton Strategies which so far has not been tested at a general election.

The one that we see the most of is the top pollster from GE2019, Opinium, with its weekly surveys for the Observer and is showing a 4% CON lead. The tightest since the start of June has been the 3% CON lead with Deltapoll.

Although Starmer has been doing very well in comparison with Johnson in the leader ratings he must surely be hoping for a LAB lead soon. I find it odd that the latest Opinium approval ratings had Johnson a net 27% behind Starmer while the voting numbers still had CON 4% ahead. I cannot recall a similar divergence in the past.

Mike Smithson


Why Starmer is not going to let go of the issue of Johnson not acting quickly enough on COVID19

Monday, July 6th, 2020

One of the great sources of analysis during the pandemic has been on Radio 4’s “More of Less” programme which each week subjects big issues in the news to statistical analysis. It has just finished its current series and its last programme was basically a summation of how the pandemic evolved. It is well worth listening to here .

Inevitably the issue of timing of the government’s decision in March is subject to scrutiny and a clear view is developing that the overall death toll could have been reduced if Johnson and his ministers had acted earlier. Even though the next general election might not take place until May 2024 the scale of this crisis and the likely examination of the government’s action by the inevitable inquiry will keep this issue alive.

The latest Opinium poll carries a question that it has asked before on whether voters think that Johnson should have acted earlier and the responses from different segments are in the chart above. The numbers suggest that Johnson has a problem while for Starmer this is an opportunity.

Interestingly the best segment from Johnson’s point of view are those in the sample who rated him as the Best PM over Starmer. Even then only 50% said he had got this right.

Mike Smithson


The end of the honeymoon

Saturday, July 4th, 2020
Wikipedia polling chart since GE2019

At some point LAB will take the lead

Labour has not led in any GB poll since 28 July 2019, four days after Boris Johnson succeeded Theresa May as Prime Minister. They did so with a share of just 30%: one more than the Tories and one less than the combined Lib Dem and Brexit Party shares. It seems a lot more than eleven months ago.

Clearly, a huge amount has changed since then but the Tory position at the top of the polls has not. This isn’t all that unusual. Prime ministers usually give their parties a boost on taking office, and parties winning election – especially when they’ve gained seats in doing so – will tend to extend their lead in the aftermath too, which is what happened.

Following the December election, every poll through to late May gave the Tories a double-digit lead, with all bar one of them in excess of the 11.8% advantage at the election. At the peak, in late-March through to mid-April, in the period immediately after the Coronavirus lockdown, the Conservatives routinely led by more than 20% and four times hit 26%.

We should note in passing that Keir Starmer’s election as Labour leader on April 4 didn’t make any great impact on those numbers. Labour’s problems clearly ran well beyond Jeremy Corbyn as an individual and the previous five years have deeply tainted their brand as a party.

From there though, the decline in the Con lead has been marked, falling around 15% in May: from about 20 down to five per cent. There are of course multiple causes of that; Cummings’ bolt to Durham, the Covid-19 death figures, Starmer demonstrably getting a grip on Labour, to name but three. The question now ought to be how long it will be before Labour secure a poll lead?

In truth, it oughtn’t be that long, for three reasons. Firstly, history. We’re approaching seven months since the election. That’s usually around the timespan by which the government falls behind in the polls. Clearly, each parliament is different, faces unique challenges and opts to take different policy choices, and this one has been more unusual than most. Even so, a glance at the record shows a good degree of consistency. These are the periods before the opposition first recorded a lead since 1979:

1979: 7 weeks (first poll of the parliament)
1983: 8 months
1987: 10 months
1992: 3½ months
1997: 3 years, 4 months
2001: 2 years
2005: 7 months
2010: 4½ months
2015: 10 months
2017: 2 days

Obviously, there’s quite a spread there but of those 10 occurrences, four land in the 7-10 month frame, with four below that and two above it. Notably, the four shortest honeymoons were for four of the five smallest majorities (including minority / coalition governments) – 2015 being the exception, another example of Corbyn underperformance. Similarly, the two longest honeymoons were for the government that won the two biggest landslides. On that basis, Johnson ought to be looking at getting on for a year.

However, he might not get there, not least for the second reason: margin of error. The average poll lead is now small enough that it might well only take one outlier to eliminate it entirely. Outliers are, by their nature, rare – and twice as rare for them to be in a particular direction i.e. in this case, pro-Lab. Still, there’s no saying when chance might throw one in.

However, it’s the third factor which is the most potent: political fundamentals. Keir Starmer has a considerable lead over Boris Johnson in terms of favourability, and is more-or-less level in ‘Best PM’ rating. Labour, as mentioned earlier, still has some way to go to win back the public’s trust but against that, the Conservatives have overseen a response to Covid-19 that’s been poor by international standards, and the economic effects of the lockdown are beginning to be felt on a wide spectrum, with job losses being announced regularly. With Covid-19 cases still in the mid-hundreds per day and local lockdowns still a risk, the chances of a V-shaped recovery seem low as even those businesses and other organisations which are in a position to begin recruiting again are likely to be wary in doing so.

The government does still have things going for it politically. The coalition it put together on Brexit isn’t likely to fracture too much while Brexit remains a live issue, as it will through to the end of the year at least. Also, there’s no obvious challenger on the right. The Conservatives retain a much higher floor to their support than they did in 2019.

All the same, I reckon that chances are that Labour will score its first poll lead in around a year before the beginning of September.

David Herdson


Perhaps pollsters should start weighting for those who have had Covid-19 or know someone who has had from Covid-19

Sunday, June 21st, 2020

This morning The Sunday Times have done some excellent analysis based on figures from the gold standard Office of National Statistics that shows

While the virus threat is receding in London and other urban centres, our research shows many of the 44 parliamentary seats in the north and the Midlands that switched from Labour to Conservative last year are suffering an above-average mortality rate. The red wall has become a blue dilemma for a prime minister determined to make last year’s breakthrough permanent.

In England and Wales the overall death rate to the end of last month was 78.9 per 100,000 people. The 44 constituencies that helped propel Johnson into Downing Street, however, had a rate of 87.7, with 14 of them recording more than 100 deaths per 100,000 people.

I’ve been wondering for a while whether the pollsters need to weight for people who have had Covid-19 or know someone infected with this disease. Simple intuition tells me that people who have been impacted by Covid-19 are likely to be more unfavourable towards Boris Johnson and the government whilst the opposite is more likely for those not impacted by Covid-19,

This has been driven by some polling from America I saw a few months ago which showed Donald Trump’s ratings doing much worse with amongst people who knew somebody infected with Covid-19, independent/swing voters seemed to have the biggest swing.

There is the possibility that I maybe over analysing things. With, inter alia, the lower paid and BAME communities being disproportionately hit by this accursed disease, groups who are unlikely to vote Conservative, then perhaps the Conservatives and Boris Johnson are doing better than the polling indicates.

With over 300,000 people testing positive in the UK so far and the experts thinking the true number is ten times that then if they know around ten different people we’re looking at 3 million people impacted, for comparison the Conservative lead in the popular vote in December’s election was just under 3.7 million that’s a fairly substantial constituency.



Johnson’s Tories still ahead with Opinium but he’s a net 30% behind in the approval ratings

Saturday, June 13th, 2020


The first phone poll since March has CON 5% ahead but Starmer with a net 32% lead on leader satisfaction

Friday, June 12th, 2020

Starmer’s numbers are the best for an opposition leader since Blair

Today we heard something of a pulling rarity – a phone survey from Ipsos Mori – the first since March well before starmer became Labour leader. The firm is the only one still doing national phone political surveys which cost a huge amount more than online polling.

One of the things I love about them is that it has been going for so long and maintains records going back to before Mrs Thatcher became Prime Minister. The firm has a wonderful consistency in the questions it asks so we can examine historical trends – hence the chart above with data on opposition meaders going back 41 years.

Although Starmer has been doing well in just about all the leader ratings that have been published which has yet to be reflected in the voting intention polls. In the UK it is the voting numbers that have the biggest impact.

Mike Smithson


A new way of showing how LAB and CON are viewed

Wednesday, June 10th, 2020

This looks quite interesting – trying to assess how the main two parties are viewed by looking at responses to the same series of YouGov questions over a period of time. This has been all dressed up and called the Valence Index.

No doubt we’ll see more of this in the coming months but it will be a long time before we can test this in a real election.

One thing that strikes me is that there are more don’t knows on Labour than the Tories that might skew this a touch. Thus on the competence question there were 74 responces for the Tories and 66 for LAB.

The strong or weak question is clearly going to be more helpful to the incumbent because it has the power to do things. Oppositions don’t and the next general election is due to take place on May 2nd 2024.

Mike Smithson


YouGov on the Bristol statue downing and the Prince Andrew/Epstein connection

Monday, June 8th, 2020