Dave’s new best friends

Prospect Magazine

Dave’s new best friends


Who will really win the election for Cameron? Backed by exclusive research for Prospect, here are the seven things you need to know

The bombardment has started, and the air is already thick with claims, polls and analyses. To arm yourself against error and ignorance, here are seven pillars of electoral wisdom for the 2010 general election.

1. To win, the Tories must carry the north, not just the south

It is widely believed that Conservative support in Scotland and the north has haemorrhaged, yet that David Cameron must do particularly well in the northern half of Britain to win a convincing majority. Neither fact is entirely true.

Across Britain, Tory support was 10 percentage points lower in 2005 (33 per cent) than in 1992 (43 per cent). Over the same period, Conservative support also fell in northern Britain: in Scotland, by 10 points; the northeast, 14; the northwest 9; and Yorkshire and Humberside, 9. Only the fall in the northeast, the smallest region, stands out. So while the Tories have ceded ground in the north, this began in the 1960s and 1970s when the Tory Protestant working-class vote declined in places like Glasgow and Liverpool, as class overtook religion as the most important issue for voters. This trend has been exacerbated by a growing economic divergence between north and south. Nonetheless, the Tories won their largest (1983) and least expected (1992) victories after this north-south divide opened up. And while they must make gains in the north to win, there are plenty of seats within reach. Thirty-four Labour seats in England’s three northern regions would fall to the Tories on an 8 per cent swing. If the national tide flows to David Cameron, enough northern seats will change hands to give him victory.

The Tories’ larger problem is that they need a lead of around 10 percentage points in the popular vote for Cameron to secure an overall majority of just one. Compare this with Labour in 2005, which enjoyed a comfortable majority with just a 3 percentage point lead. Why? The main reason is that, even after the latest boundary changes, Labour seats tend to have fewer voters, and lower turnouts, than Tory ones, so Labour on average needs fewer votes to win a seat. Also, the last three elections have seen widespread anti-Tory tactical voting. In 2010, the first factor will still apply; one of the more intriguing questions is whether anti-Tory tactical voting will unwind. Even so, the electoral geography is stacked against the Conservatives.

A smart approach to the 2010 election would divide Britain into four zones: the west, south, Midlands and north (see map, overleaf). Adjusting 2005’s result for changed boundaries, Labour’s existing 66-seat majority falls to a notional lead of only 48. So if Cameron wins just 24 seats, he knocks off Labour’s overall advantage. Here, 16 of these 24 “super-marginal” Lab-Con seats (those contests where the Conservatives are in close second place, coloured blue on our map) are in the south. Many have stations on the Thameslink rail line from Bedford to Brighton. The commuters who live in such places, who we call “Thameslink voters,” share the same broad social mix of those who will decide the election nationwide (as Sam Knight explains, p36). But if the Tories are to become the largest party, they must take a second tranche of 40 or so seats (coloured purple), many of which are in the Midlands. And to win an overall majority they must prevail in a third tranche of contests (coloured red), mostly in the north.

Why not include Scotland in this four-way grouping? Scottish voters vote differently from the rest of Britain, in part because of the SNP. But while the 2010 result has important implications for Scottish politics, it actually matters little for the arithmetic at Westminster. The Tories hold just one Scottish seat. If they do incredibly well, they could win another seven, including the long shot of unseating Jim Murphy, the Scottish secretary, in Renfrewshire East. But gains of only one or two are more likely. The Tories need 116 gains to win an overall majority—making Scotland just a sideshow.

2. The election won’t be decided by a few thousand votes in a handful of seats

After an election some expert always says something like: “Had 10,000 voters in 20 seats voted another way, the result would have been different.” There is a narrow arithmetical truth in this. If a party wins a seat by 1,000 votes then it would have lost had 501 switched to the second-placed party. Apply this logic to the 20 most marginal seats, and you could argue that just 10,000 votes matter, and the other 30m don’t. But there are two problems with the proposition. The first is philosophical. Imagine someone swimming from Dover to Calais. If she stops 10 metres short she fails to finish. But that is also true of every 10-metre stretch of her swim. To succeed, she needs to swim all 34km. By the same logic, to win a seat, a candidate needs around 20,000 votes; to single out the tiny number that comprises half-the-majority-plus-one is absurd.

Second, such calculations can be done only retrospectively. In 2005, for instance, Bill Rammell won the southeast seat of Harlow by only 97 votes. Robert Halfon, the losing Conservative, might have kicked himself, thinking, “If only we had persuaded 49 more Labour voters to support me instead, I would have won.” But he fought the campaign not knowing those numbers, needing to overcome Rammell’s 2001 majority of 5,228. He had to keep the Tory base solid, get his supporters to vote, and persuade as many others to vote for him as possible. Suppose he had said to his campaign team at the outset: “I have had a vision of the result: we shall fall 97 votes short. Let’s junk our strategy, and find only 49 extra Labour supporters to vote for us.” They would have laughed at him.

Nationally, the story is more complex. Most MPs are safe. The real battleground comprises at most a third of the 650 seats, but this still means many millions of votes “really” matter. Even then, surprises can happen. During the 1997 election, the BBC produced a “bible” containing detailed constituency information. What did Crosby, Hove and Harrow West have in common? The “bible” tagged them all as “very safe” Conservative seats—and all fell to Labour. In more recent years, Labour too has suffered losses in such strongholds as Blaenau Gwent and Bethnal Green and Bow. Parties must nurture their heartlands, not just protect their marginals.

3. Essex man and Worcester woman don’t decide elections

It is hard for journalists to resist finding “typical” voters in a marginal seat, and asking their views. Such exchanges can illuminate how real people view the parties and their leaders. But problems come when such encounters are invested with cod sociological authority, especially when a handful of voters are labelled as holding the key to Downing Street. This is invariably nonsense. Take “Worcester woman.” She was discovered by the Sunday Times shortly before the 1997 election. A front page story, complete with a poll, explained how Labour was doing well among women in the west Midlands, and dramatised the finding by putting “Worcester woman” in the headline. The label stuck, even though not one of the poll’s respondents actually lived in Worcester. Worse, the subsample of west Midlands women was too small to justify even the basic claim that they were unusually pro-Labour.

In general, it is rare for particular demographic groups to move very differently from the electorate as a whole. There are exceptions: both Muslims and students deserted Labour in 2005 over Iraq and tuition fees respectively; as a result, the party lost a handful of seats it might otherwise have retained. But movements from one election to another are notable more for their uniformity than their variation. When Tony Blair swept to power in 1997, Labour gained ground among rich and poor, men and women, young and old. If David Cameron is to win the coming election, he, too, must win converts in nearly every demographic group.

4. The Lib Dems could deny David Cameron victory

The fourth of our election battlegrounds on the map sits mostly in the west of England and Wales, where the majority of Tory/Lib-Dem contests can be found (coloured yellow). When the Tories last won an election, in 1992, the Liberal Democrats took 20 MPs. In 2005 they won 62. Almost all of these extra 42 are in natural Tory territory. Can Cameron win them back? Probably not. Look back to 1979, the last time the Tories won power from opposition. Then their share of the national vote climbed eight points, while the Liberal vote fell by four (figures roughly similar to today’s polls). Had these movements occurred in each of the 13 seats then being defended by the Liberals, seven would have fallen to the Tories. In fact only three did. In the other four, the Liberals gained votes, keeping the Tories at bay. The same may happen this year. National polls imply that Cameron should win around 20 seats from the Lib Dems; in practice he may win far fewer. Indeed, the Lib Dems may capture one or two Tory seats. Boundary changes have made Solihull a notional Tory marginal, but the incumbent (on the old boundaries) is Lib Dem MP Lorely Burt—and I would not bet too much against a “Lib Dem gain” here, come election night. The Tories face another handicap in the 17 seats won by Margaret Thatcher where they have fallen into third place. Nine of these were won off them by Labour, but have in turn been taken by the Lib Dems. Almost all are out of reach for the Tories this time round. All in all, it will be surprising if net Tory gains from the Lib Dems climb above single digits.

More generally, Cameron’s task is made much harder by Britain’s unusually large block of third and minor party MPs. In 1979 there were just 28 such MPs, meaning Thatcher was able to govern comfortably having won just 71 more Tory than Labour MPs. But by 2005, that 28 had more than trebled to become 92. So this year, if the Tories manage to lead Labour by 71—on the face of it a good result, given Labour’s lead of 158 after 2005—Cameron will almost certainly not secure an overall majority. More significantly, as long as at least 70 MPs belong to a mix of Lib Dems and minor parties, future elections could throw up hung parliaments too.

5. Issues don’t win elections; valence does

Cast your mind back to the 2000 Romsey byelection, in Hampshire. It was among the safest Tory seats. At the time immigration was the dominant issue, especially as dramatic news reports showed asylum seekers at the Sangatte camp, near Calais, attempting to jump onto trains entering the Channel tunnel. The Tories sought to exploit this in their campaign. Yet the Lib Dems, whose relatively liberal immigration policies were backed by few voters, won easily.

Why? Because “valence” matters. This unusual term is used by political scientists to mean the way voters make judgements about who to vote for. It brings together a number of issues, including the way people think about the competence, character, authenticity and performance of their potential leaders. To understand “valence voters,” contrast them with “positional voters.” If you feel strongly that greater private involvement in healthcare is a good thing—or bad—you have a positional view of the NHS. If you don’t feel strongly either way, or don’t know much about it, but simply want a prime minister you can trust and a government that will ensure you are treated promptly and effectively when you fall ill, you are a valence voter. Experts and political nerds hold strong positional views about many aspects of public policy. But most voters—and a large majority of floating voters—don’t. They care mostly whether politicians are decent, honest, capable and on their side.

Valence theory matters because, for instance, any benefit the Tories gained in Romsey by a tough stance on immigration was massively outweighed by the view that they were cynically exploiting the issue for electoral gain. And what was true in Romsey is true about virtually every issue nationally today. Thatcher won the valence war with successive Labour leaders in the 1980s (“she’s strong, knows what she wants, and will sort Britain out”). John Major then lost it in the 1990s (“he’s weak and leads a divided, sleazy and incompetent party”). Now, the contest between Gordon Brown and Cameron will be decided by the valence judgements of millions of voters who have yet to decide their vote. The “issues” that will determine their choice will be perceptions of Brown’s competence, honesty and dour demeanour (factors that could be either positive—he is a serious man who has steered us successfully through the worst economic storms for 80 years—or negative—he is a sour-faced spendthrift who belongs to the past, not the future) versus perceptions of David Cameron’s freshness, charisma and competence, and whether voters think he understands the lives of normal voters.

6. Polls have improved, but aren’t perfect

Pollsters were traumatised by their failure to predict the Conservative victory in 1992. All have since changed their methods, albeit in different ways; and new companies (like the one for which I work, YouGov) have sprung up. Even so, three reasons suggest the polls might not be right this time either. First, telephone polls conducted by ICM, Mori, NOP and Populus in 2005 appear to have retained a slight Labour bias. This meant that while Labour’s support remained stable during the campaign, a majority of such polls overstated the party’s final vote-share by three points or more, probably because some voters were still reluctant to admit to a stranger that they were going to vote Conservative. Things may be different this time; but we will not be able to tell until afterwards.

Second, even well-conducted polls are subject to sampling error. Statisticians think a poll result is perfectly valid if it is within two points of the actual result, but statistical accuracy is not the same as political accuracy. Even one-point errors can change the political significance of the election. Suppose the final polls indicate: Conservative 40 per cent, Labour 30 per cent. If the result is 39-31, Britain is likely to have a hung parliament; but if it is 41-29, Cameron is likely to secure a clear majority. Finally, polls measure votes, but it is seats that matter, and votes translate into seats in unpredictable ways. In February 1974, the final polls pointed to a narrow Conservative lead in the popular vote, and headlines predicted victory for Edward Heath. The polls were right about who would win the popular vote—but Labour emerged slightly ahead on seats and Heath had to resign. The polls were blamed unfairly for getting the election wrong. It is best to follow the dictum of Adlai Stevenson, Democratic candidate for the US presidency in the 1950s: “Polls should be taken but not inhaled.”

7. Campaigns don’t win elections, but stuff happens

In general, election campaigns produce more noise than electoral movement. Most voters make up their mind before the campaigns start; and those who wait tend to divide fairly evenly between the main contestants. However, there are exceptions. During the past decade two European elections have been upended by unexpected events. In 2002, Germany’s Gerhard Schröder was heading for defeat, his administration mired in scandal and economic failure. But just before the elections, the river Elbe burst its banks, and much of eastern Germany was stricken by floods. Schröder’s decisive response revived his popularity; his red-green coalition won a narrow majority. Eighteen months later, Spain’s centre-right government looked set to be re-elected when, three days before polling day, a series of bombs exploded in Madrid’s train network, killing 191 people. Initially the government blamed Basque separatists, but it became clear that al Qaeda supporters were responsible. The opposition socialists benefitted from the government’s bungled response and won the election. It is perfectly possible that some deus ex machina—hopefully less calamitous than a flood or a bomb—could alter the course of this election, and so do enough to produce either a Tory landslide or a fourth Labour victory. There is, however, one “known unknown”: the television debates between the party leaders. If any of them drops a mighty clanger, he will suffer.


  1. February 27, 2010


    Should we allow opinion polls to be published during a General Election campaign? Discuss.

    No we definitely should not.

  2. March 11, 2010


    We now have the date of the budget and speculation that an election will be held on 06 May 2010 following the “wash up”. Having been subjected to wall-to-wall opinion polls since Gordon Brown was accused of bullying and then appeared at Chilcot, I have made certain decisions. One of them is that I will vote according to my conscience.

    “In general, election campaigns produce more noise than electoral movement. Most voters make up their mind before the campaigns start..” – as per heading 7 above.

    Given the egg all over the face of the polls in 1992, I will take that as a suggestion by Peter Kellner that we should not bother with elections and just let him and the opinion polls decide it. The changes of method per heading 6 above do not alter the fundamental problem. There are getting on for 40 million electors and each poll reports the opinion of less than 1000.

    Pollsters are to elections and democracy what bankers have become to the economy. No poll should be published unless it contains a complete disclosure of what Peter Kellner earns from publishing it, who pays that money and why I have never been asked for my opinion, why nobody I know has ever been asked for their opinion and why all the people I know cannot find anyone who has ever been asked for their opinion.

    Immediate opinion poll licensing is required. All participants in a survey must be named before the poll can be published. Pollster remuneration must be disclosed and all interests likely to influence the poll must be declared by participants and the pollsters. 3 years’ audited accounts and enhanced CRB disclosure must be provided by all pollsters and participants in polls stating any incentive payments given to participate. This is necessary now in case we change the voting system.

    The encouragement to tactical voting that polls corruptly offer is now only happening under the block vote. If the referendum to change to the alternative vote goes ahead and is passed, the abuses will become more legion and polls should then be banned by law completely.

Leave a comment


Peter Kellner

Peter Kellner
Peter Kellner is president of YouGov and author of “Democracy: 1,000 years in pursuit of British Liberty” (Mainstream) 

Share this

Most Read

Prospect Buzz

  • Prospect's masterful crossword setter Didymus gets a shout-out in the Guardian
  • The Telegraph reports on Nigel Farage's article on Lords reform
  • Prospect writer Mark Kitto is profiled in the New York Times

Prospect Reads

  • Do China’s youth care about politics? asks Alec Ash
  • Joanna Biggs on Facebook and feminism
  • Boris Berezosky was a brilliant man, says Keith Gessen—but he nearly destroyed Russia