Simply the best: let’s talk about those ‘Best Prime Minister’ polls

Over the past three years, those of us who support Jeremy Corbyn’s leadership of the Labour Party have heard more or less every criticism under the sun. Public opinion polls, and Labour’s performance in them, have been one of the many criticisms heaped upon Corbyn by the commentariat and the Labour right. Every Conservative lead has been instantly trumpeted as the final ‘proof’ of Corbyn’s unelectability, even after Corbyn turned a 25pt Conservative poll lead into a 2pt lead in the space of two months back in 2017. But even when the standard voting intention polls look fine for Labour (or even good), Corbyn’s critics always have one reliable poll question that they cite over and over again – “Who would make the best Prime Minister?”.

So let’s talk a bit about this question, and how it might not mean as much as Corbyn’s critics think. 

Background 

Since Theresa May become PM in July 2016, pollsters have begun asking this question more and more. It is now a standard feature of polls published by YouGov and Opinium, as well as occasionally being asked by other pollsters. It’s a simple enough question: pollsters ask voters to pick the person they think would make the “best” Prime Minister, asking them to pick between May and Corbyn. Most pollsters include ‘don’t know’ as an option; some add ‘neither’ or ‘none’ as another option.

So what sort of results do these polls show?

Well, in July 2016, when May became Prime Minister, here’s what the polls said on average: 

Theresa May – 52%
Jeremy Corbyn – 18%
Don’t Know/Neither – 31% 

Meanwhile, here’s the average result for September 2018: 

Theresa May – 35%
Jeremy Corbyn – 26%
Don’t Know/Neither – 39% 

The table below shows the average result for each party leader on this question since May became Prime Minister in July 2016.

So, why does this matter? 

Since Corbyn consistently polls behind May on this question, these polls are constantly cited by Corbyn’s critics as proof that Corbyn is an electoral liability, and that it is him personally that the electorate do not like. Thus, they say, if we got rid of Corbyn, Labour would skyrocket in support and be 20pts ahead in the polls.

Yet even setting aside the fact that Corbyn has shown himself to be fully capable of turning around public opinion (between April and June 2017, his standing in the ‘best PM’ polls improved by 20pts), this argument is weak. Here's why.

Are pollsters really asking the right question? 

All of the various poll companies have a slightly different way of asking the ‘best PM’ question:

• YouGov, Survation and Lord Ashcroft ask: “Which of the following do you think would make the best Prime Minister?”
• Opinium ask: “Which, if any, of the following people do you think would be the best Prime Minister?”
• Ipsos MORI ask: “Who do you think would make the most capable Prime Minister?”
• Kantar Public ask: “Who do you think would make the best leader for Britain?”
• And finally, BMG Research ask: “Who would you prefer to see as the next Prime Minister?”

There is a notable exception in that list: uniquely, BMG Research asks voters to tell them who they want to be Prime Minister, not just who they think would be the “best” at the job. All the other pollsters ask voters to express a judgement on which party leader would perform best at the role of Prime Minister: looking the part, acting like a PM, going on all the international trips abroad, attending summits, handling diplomacy – things like that. But that’s not a value judgement - it’s not even a question about competence. 

It’s basically asking people: who do you think would be better at being Prime Minister? The person who is Prime Minister, or somebody who isn’t? The answer is pretty obvious to most people.

This difference in wording is reflected in the data. BMG consistently shows poorer results for Theresa May than either YouGov or Opinium, the only other pollsters to publish a monthly figure for 'best PM'. Sometimes BMG shows May doing 10pts worse than YouGov. That’s not just a difference in sampling or turnout weightings – that’s the consequence of asking people a different kind of question. 

When you ask voters who would make the “best” Prime Minister, more people say May. When you ask voters who they want to be Prime Minister, May’s commanding lead vanishes like a puff of smoke. That’s worth thinking about.
 

Does any of this even matter? 

The reason why Corbyn’s critics keep pointing to these polls is because they think the public will refuse to vote Labour because of the leadership, and that the question of who makes the “best” Prime Minister is very important in determining voting intention. But survey data suggests that this isn’t even true.

Following the 2017 general election, YouGov asked voters to tell them the main reason why they backed Labour or the Conservatives in the general election. 17% of Labour voters cited their support of Jeremy Corbyn or their opposition to Theresa May as a reason. Meanwhile, 19% of Conservatives cited their support of Theresa May or their opposition to Corbyn as a reason.

In other words, for over 80% of the two major parties’ voters, the merits or weaknesses of the party leaders were less important than issues like education, Brexit or policy generally. And this remains the case today. In a recent YouGov poll, just 7% of voters said that “Which party's leader would make the best Prime Minister” was the most important issue in deciding their vote. To over 90% of voters (and over 95% of Labour voters) there are much more important things on their mind when they go to the polling station.

So in the end, whilst it’s interesting to note the differences between BMG’s polling and YouGov’s polling, to most voters the question of who makes the “best” Prime Minister is pretty unimportant. If Corbyn’s critics want to become relevant in the Labour Party again, they really need to stop obsessing over opinion polls and focus on developing an actual policy programme – such as Corbyn’s promise to nationalise the water industry, a policy which one poll suggests is backed by 59% of the British public. That’s one poll that you won't hear about from Corbyn's critics.

--

If you enjoy Stats for Lefties, please become a Patreon subscriber from as little as $3 a month! You'll get exclusive articles and resources, and you'll help fund us sustainably so we can do cool things like podcasts and videos. patreon.com/Stats4Lefties

The polls aren’t moving. What does that mean?

Given that the British government has been lurching from crisis to crisis, and that Labour leader Jeremy Corbyn has faced mountains of negative media coverage, you might expect public opinion to have shifted dramatically back and forth depending upon the latest news. After all, between April 2017 and the post-election period of June 2017, Labour’s average poll rating rose from 27% to 44% - an increase of 18 percentage points (5.6m votes) in just two months. We are not exactly living in an era of political stability.

And yet, since Theresa May lost her majority in June 2017, neither party has managed to lead the other by an average of more than 3pts, as shown in the table below.

Indeed, most of the time the parties have been ‘statistically tied’. This means that whilst one party may be in the ‘lead’, that lead is so small (just 1pt according to the latest poll average) that we really don’t know what public opinion is. It could, within a reasonable “margin of error”, go either way. The British Polling Council explains it like this: 

“All polls are subject to a wide range of potential sources of error. On the basis of the historical record of the polls at recent general elections, there is a 9 in 10 chance that the true value of a party’s support lies within 4 points of the estimates provided by this poll, and a 2 in 3 chance that they lie within 2 points."

In other words, if the Tories are on 39% and Labour is on 39% in a poll, that doesn’t mean both parties would be tied if an election was held tomorrow. Instead, what it tells us is either party could be ahead. With the normal margin of error on a party’s vote share being plus or minus 4pts, the Tories could really be as high as 43% - or as low as 35%.

Essentially, when any party is leading by less than 10pts, the specific numbers don’t really matter. If the polls are effectively tied, it means we really have no idea how an election tomorrow would go. The Tories could win by a big margin or get defeated decisively. Conversely, if one party is leading by a significant margin well outside the margin of error (e.g. by 25pts), they’d be very likely to win the popular vote in an election held tomorrow – probably not by the predicted margin, but they’d win it nonetheless. Bear in mind that in the final polls for the 2017 election, the Tories’ average lead was 7pts – they did indeed go on to win the popular vote, but by a much smaller margin of 2pts.

Given all of this, what can we tell from the opinion polls since June 2017? Here are some things we know for sure:

1. We don’t really know who is more popular: Labour or the Conservatives.

Sometimes political parties achieve large leads over their opposition in the polls that show unambiguously who is more popular at that moment. For example, in April 2017 the Conservatives were, on average, 18pts ahead of Labour, which would have given them a majority of 100 seats in the Commons. In one poll they had a staggering 25pt lead. Yet in the 15 months since the 2017 election, the biggest average lead achieved by either major party has been just 3pts (usually just 1pt!), and no party has been projected to win a majority on the 650-seat boundaries in any month since June. The maximum lead by any party in an individual poll has been 8pts, and that has only happened once or twice.

Individual polls have shown varied results when it comes to the major parties, sometimes to the point of being absurdly different. In March, Labour was leading by 7pts with Survation but trailing by 3pts with YouGov in the same week. Recent polls have varied from a 4pt lead for Labour to a 5pt lead for the Conservatives. The only story being told by the polls is that either major party could win the popular vote if an election was held tomorrow.

2. The major parties have leaked a little support to minor parties, but not much.

In the 2017 general election, the two biggest parties won a combined 84.5% of the popular vote in Great Britain, the biggest combined share in over 40 years. The Lib Dems went backwards, achieving their lowest share of the vote since 1959 (7.6%), UKIP collapsed from 13% to 2% and the Green Party lost half their vote. The SNP, meanwhile, retained a majority of Scottish seats, but lost 11pts in vote share and 21 seats.

Since the election, minor parties’ support has slightly increased. The combined vote share of the two major parties is down to 77% (-8%), but no particular minor party has benefitted from this. The Lib Dems and Greens have each gained 1pt since the general election, barely edging up at all. UKIP has gained 3pts in the polls and is currently averaging at 5% of the vote, whilst the SNP are projected to gain just 4 MPs. The clearest story from the polls is that minor parties are continuing to struggle in a politically polarised country, and the Lib Dems in particular are still unable to poll any higher than 10%.

3. The outcome in terms of seats is very uncertain.

One element of the 2017 election that went uncommented upon was that it was really, really close. The gap between the two biggest parties in the national popular vote (42.4% to 40.0%) was the smallest since October 1974, whilst at a constituency level, a total of 96 seats were decided by a margin of less than 5pts - 40 Conservative, 31 Labour, 15 SNP, 4 Lib Dem, 2 Plaid Cymru, 2 Sinn Fein and 1 DUP. This leaves us with a high degree of uncertainty in assessing how many seats each party will win. A small swing to one party or the other could result in big gains, or a strong ground campaign could see parties retain close seats that they might otherwise have lost.

For example, using the standard ‘universal swing’ method (applying the national ‘swing’ to all seats), the September poll average would lead to the following results:

CON 39% (-5%), 305 seats (-12)
LAB 38% (-3%), 265 seats (+3)
LIB 9% (+1%), 18 seats (+6)
UKIP 5% (+3%), 0 seats (n/c)
SNP 3% (n/c), 39 seats (+4)
GRN 3% (+1%), 1 seats (n/c)
OTH 1% (n/c), 22 seats (-1)

On the face of it, first-past-the-post would give the Tories a big bonus, as their 1pt lead in the popular vote would equate to a 40-seat lead (a lead of 6.1pts in terms of seats). But if, instead, Labour beat the Tories by 1pt in the popular vote, the Tories would win 286 seats to Labour’s 285. In other words, a tiny shift in votes would cause a dramatic shift in seats, from a decisive Tory lead to a dead heat.

Thus, given that all polls have a margin of error of up to 4pts – and that a shift of just 1pt can now shift 20 seats – anyone pronouncing with absolute certainty that Labour is dead in the water really doesn’t have much of a basis for that.

So, ultimately, the opinion polls are sending us a clear message: if an election was held tomorrow, it would be wide open, and anybody could win. Both major parties are polling at historically high levels, and are thus virtually tied, whilst the minor parties continue to struggle. In the past 15 months, nothing has broken this deadlock to any significant degree. In all likelihood, until something dramatic happens (such as a snap election, the rejection of May’s Brexit deal or a leadership challenge to either May or Corbyn) this deadlock will continue until the next general election – and then the polls will actually start to mean something.

Until that election comes, these polls are largely a curiosity. Nobody should be panicking about them or obsessing over every little shift and subsample. They are interesting for geeks like me, but the main thing people need to know is this: in 2017, Theresa May began the campaign with a 25pt lead in the polls, and by the time people actually voted, she only won by 2pts. If that doesn’t show you how you should never shape your politics around what the polls say, I don’t know what would.

What do the proposed Labour leadership election rules mean?

Update on 10.56pm, 22nd September: Labour's NEC have agreed new leadership rules, but they differ from the below in a vital way. Rather than requiring 10% of MPs/MEPs AND 5% of CLPs AND Unions representing 5% of affiliate members, the NEC's proposal would require:

• 10% of the PLP

Plus EITHER:

• 5% of CLPs

• 5% of affiliated organisations, at least 2 of which must be Trade Unions who represent at least 5% of the affiliated membership (currently estimated at 3.9m).

The important word here is "either". The upshot is that the trade unions would not have the ability to solely veto a candidate; candidates can be nominated if they gather the backing of 25 MPs, plus either 30 CLPs or one big trade union.

These rules would have allowed Andy Burnham to remain on the ballot in 2010, with Ed Balls and Diane Abbott still excluded. Yvette Cooper would have made the ballot in 2015, as she recieved 25% of PLP nominations plus 17% of CLPs. Liz Kendall, however, would have been excluded.

-----

Within the last few days, more details have emerged about Labour’s soon-to-be-finalised Democracy Review. It is looking very likely that one of the proposals presented to Labour Conference by the National Executive (NEC) will be a change to the rules for electing the Labour Party Leader. Specifically, the changes being proposed would affect how candidates are nominated (the voting system would apparently remain unchanged). 

The current system 

Under the current rules, in the event of a vacancy, Labour MPs can stand to be Leader of the Labour Party if they gather written nominations from 10% of MPs/MEPs. As Labour currently has 257 MPs and 20 MEPs, a candidate would need to gather nominations from 28 MPs/MEPs. This number has shifted over the years (12% in 2010, 15% in 2015, reduced to 10% in 2017) but the key point is that the Parliamentary Labour Party (PLP) has retained the absolute sole right to nominate candidates for Leader, with other sections of the party only permitted to offer “supporting” (i.e., non-binding) nominations.  

The proposed system 

The nomination system that will be debated by the NEC on Saturday would retain the requirement for candidates to be nominated by at least 10% of the PLP. However, it also adds two new requirements in addition to this PLP threshold. To be on the ballot, candidates would have to receive nominations from at least:

• 10% of MPs/MEPs
• 5% of Constituency Labour Parties (CLPs)
• 5% of affiliated organisations, at least 2 of which must be Trade Unions who represent at least 5% of the affiliated membership (currently estimated at 3.9m).

There are only 4 out of 14 affiliated Unions who represent at least 5% of this 3.9m total: UNISON (34.6%), Unite (31.9%), GMB (15.7%) and USDAW (10.9%). Together these four Unions represent 93% of Labour’s affiliated Union members. Thus, in practice, leadership candidates would need the support of at least one of the big four (plus any other union, large or small) in order to be on the ballot. 

The consequences 

The possible effects of this proposed system have been raised most strongly by the Labour left, who believe that it effectively rules out the possibility of there being a successor to Corbyn who shares his political outlook. It is worth taking a look at the effect on previous elections.

Table showing nominations by affiliates, CLPs and the PLP/EPLP in 2010, 2015 and 2016. 'Union %' indicates the percentage of affiliated members represented by the unions who nominated that candidate.

As the table above shows, the proposed rules would impose a much more prohibitive threshold upon candidates. Neither Ed Balls, Andy Burnham nor Diane Abbott would have made the ballot in 2010, all falling at different hurdles. Yvette Cooper and Liz Kendall would both have been blocked from the ballot in 2015, with Cooper failing to pass the Union threshold, as she gathered support from Unions representing a mere 1.3% of affiliated members. Even David Miliband’s position in 2010 would have been precarious; he was only nominated by two trade unions, and if he had failed to gather that second union nomination, then under the proposed rules Ed Miliband would have been elected unopposed!

Conclusion

Some critics have depicted the changes as giving the PLP a veto over candidates, but this is a mischaracterisation. In fact, the biggest beneficiaries by far would be the big trade unions. The proposed changes would give the five biggest trade unions – UNISON, Unite, GMB, USDAW and CWU – an effective veto on who is allowed onto the ballot for Labour leadership elections, as together they represent 97% of affiliated members.

The question is – what does this mean for Corbynism? After all, Corbyn himself easily passed the threshold in 2015 and 2016, winning nominations from Unions representing over 70% of affiliated members; given this, the objections from the Labour left might seem confusing to the casual observer. But they are right to be concerned.

It is historically very unusual for the affiliated unions, especially the large unions, to back the Labour left’s candidate in a Labour leadership election. In 2010, Diane Abbott received nominations from unions representing just 1% of affiliated members. But in 2015, all candidates except Corbyn had made clear they did not want any support from trade unions, nor did they agree with the unions’ policy priorities. Corbyn, on the other hand, welcomed the unions’ support and his domestic and economic policy priorities dovetailed perfectly with the unions’. When combined with grassroots pressure on union leaderships from pro-Corbyn union members, this led to the unusual spectacle of the biggest unions almost all lining up behind the Labour left’s candidate. 

But with Labour’s shift to the left, and with ‘soft left’ MPs realising that their politics might actually win over the membership, we are unlikely to see a situation in which a sole Left candidate faces off against multiple centrists. More likely is a scenario in which a Left candidate, a soft left candidate and two or three centrists attempt to gather nominations. And given how high the threshold would be under these proposed rules, I think Momentum is more than justified in worrying that a candidate from the Labour left would be in real danger of not reaching the union threshold, as the big unions could (as they did in 2010) line up behind the ‘soft left’ candidate. 

So for now, the future of the Labour left rests in the hands of the NEC and Labour Party Conference.