With just under five weeks to go to the general election, who do the forecasters think will be left holding the keys to Downing Street?
The Conservatives and Labour are pretty much neck and neck.
Three new polls have come out in the last 48 hours. YouGov have the Tories and Labour tied on 35 per cent. Populus also has a dead heat, with both parties on 34 per cent.
Lord Ashcroft’s weekly poll has the Tories in front on 36 per cent and and Labour on 34 per cent.
Ukip are ahead of the Lib Dems in all three polls and Lord Ashcroft has Nick Clegg’s party in fifth place behind the Greens.
The Press Association and the BBC’s “poll of polls” have Labour and the Conservatives on 34 per cent each, with the two biggest parties gaining in support recently at the expense of the smaller ones.
Read more: Election 2015 live blog
Using the polls to predict what will actually happen on May 7 is a complicated business. Pollsters and academics have developed various mathematical models, variously using national and local polling, data from previous elections and the census.
Many models suggest that Britain is currently heading for a hung parliament, where no one party wins the 326 seats needed for an absolute majority in the House of Commons.
That suggests there will be another coalition government or a less formal arrangement where smaller parties agree to support a party that does not have an absolute majority.
Electoral Calculus predicts that Labour will win slightly more seats than the Conservatives – 284 compared to 279:
Other forecasts by the Guardian, ElectionForecast.co.uk and the New Statesman’s may2015.com site put the Conservatives slightly ahead.
An interesting outlier is Elections Etc, a forecasting model which tries to allow for the fact that there is often an inherent bias in polling.
Historically, support for the government of the day tends to rise in the run-up to elections, and the Conservatives tend to do better than opinion polls suggest.
This leads Elections Etc to conclude that the Conservatives are likely to win substantially more seats than Labour, and have a much better chance of winning a majority (16 per cent compared to 1 per cent):
But a hung parliament is still by far the most likely outcome, and David Cameron is still only marginally more likely to be prime minister – 54 compared to 46 per cent.
Electoral Calculus suggests a Labour prime minister is a much more likely outcome, assuming the SNP keep their promise not to help the Conservatives.
The SNP could wield considerable influence, according to the site, with a good chance that Alex Salmond will decide who will be the next prime minister or be deputy prime minister himself.
The question no party is keen to talk about. But it is a live possibility that even the most probable two-party coalitions will fail to win an outright majority.
In a recent Telegraph article, Martin Baxter, founder of Electoral Calculus, raises the prospect that the Conservatives could be the largest party, but might still not be able to form a majority even with the Lib Dems.
At the same time, a Labour/SNP bloc could also fall short of the magic number of 326 seats. A Conservative/SNP coalition government would be mathematically possible, but the nationalists appear to have firmly ruled it out.
So what happens next? Either a potentially multi-party coalition is formed to create a small majority, say Labour plus the SNP and the Lib Dems. Or someone could try to form a minority government – but a vote of no confidence would trigger another general election.
Or a minority government could get enough support to avoid a no confidence vote and try to struggle on, attempting to muster support from various parties for every Commons vote.
Practically speaking, David Cameron would remain prime minister while the parties negotiated with each other – as Gordon Brown did in 2010. With more parties involved, the horse-trading is likely to go on for much longer than the five days it took to form the Conservative-led coalition.
If a formal coalition government cannot be formed, it is possible that two or more parties could do a “confidence and supply” deal, where the smaller parties agree to back a minority government on key tests like the Queen’s Speech and votes of no confidence in return for some kind of concession.
Despite a surge in support for the two smaller parties in the polls, most election prediction models show that this is unlikely to mean they win substantial numbers of seats.
Electoral Calculus and ElectionForecast.co.uk predict that Nigel Farage’s party will only win one seat while the Lib Dems will hang on to many more – even if Ukip wins a bigger share of the national vote.
This is because Britain’s first-past-the-post system means support at national level does not translate evenly into victories in real constituencies.
Other sites are predicting more seats for Ukip, and most pundits admit that there is a higher level of uncertainty with Ukip due to a lack of historical data on the party.
Most of the country’s 650 parliamentary seats will not change hands on May 7. As ever, the battle will be decided in dozens of key marginal constituencies, where a small swing in support can see the seat change hands.
Warwickshire North is the Tory seat with the smallest majority – just 54 votes. And Labour veteran Glenda Jackson is clinging on in Hampstead and Kilburn with just 42 votes.
The website Democratic Dashboard lets voters find out whether the seat they live in is a safe or marginal one.
The rise of smaller parties make some seats particularly interesting. Thurrock and Great Yarmouth, both Conservative/Labour marginals, are also being targeted by Ukip.
Lord Ashcroft, the pollster and Conservative donor who has just quit the House of Lords, is one of the biggest suppliers of local polling data in key marginals.
His latest report suggests that Lib Dem leader Nick Clegg is in danger of being unseated in Sheffield Hallam, despite a hefty 15, 284 majority.
But the Lib Dems share was level or up in five out of six seats, while the Ukip share was down.
Lord Ashcroft said: “The mix of results this time round underlines the lack of any uniform swing and the hazards of trying to calculate seat numbers on the basis of national vote shares.”