Tonight’s YouGov/Sun poll has topline figures of CON 37%, LAB 43%, LDEM 8%. Yesterday’s 2 point lead looks like it was just a blip, and the 6 point leads we’ve seen from YouGov lately are restored. Tomorrow will be the first chance to see if Ed Miliband’s speech has had any impact, though lots of the fieldwork will be done before people see the news tonight or the newspaper reaction tomorrow, so worth waiting until Thursday before drawing any conclusions.

Over on ConservativeHome there are details of some polling commissioned by Lord Ashcroft (and presumably, though he has an annoying habit of not mentioning who carried out the polling, conducted by Populus). Ashcroft commissioned three parallel polls with identical questions – a telephone poll of Conservative marginals with Labour in second place, a telephone poll of Conservative marginals with Lib Dems in second place, and an online national poll.

The online national poll had topline figures of CON 37%, LAB 37%, LDEM 11%. It was conducted in late August, so at the time all the polls were showing very tight Labour leads, and is largely interesting only because if Populus did conduct it, it’s the first time we’ve seen an online poll from them.

In the marginal seats, the poll used the same trick I did back in the PoliticsHome polling of marginal seats from 2008. First it asked people their voting intention using the standard question, THEN it asked them their voting intention again saying “thinking about your own constituency and the parties and the candidates who are likely to stand there, which party’s candidate do you think you will vote for in your own constituency at the next general election?”

This makes a relatively minor difference in Con -v- Lab seats (mainly it reduces the proportion of people saying “other”, presumably because some people out there would like to vote Green or BNP or something, but don’t normally have a candidate from that party to vote for). However, it makes a major difference in Con -v- Lib Dem seats. In a normal voting intention question in Con -v- LD seats the Lib Dems are in third place on 18%, asked using the constituency specific wording they are on 31%. Labour are 7 points lower using the wording asking about people’s own constituencies.

The reason for the difference is most likely tactical considerations – people answer Labour to a normal voting intention question because that’s the party they really support, but know that they happen to live in a seat where Labour could never win, so actually vote Liberal Democrat. I found exactly the same pattern when I first asked the question in this form for PoliticsHome, but the fact that it still produces the same pattern of results is very good news for the Liberal Democrats – it suggests that in Con -v- LD seats many (but not all) Labour supporters will still vote LD tactically.

In the Con -v- Lab marginals, voting intention with changes from the general election were CON 35%(-4), LAB 44%(+8), LDEM 12%(-5). This is a swing of 6 points (the equivalent of a Labour lead of about 5 points in a national poll), so suggests Labour may be doing somewhat better in key marginal seats than in the country as a whole.

In the Con -v- LDem marginals, voting intention figures are CON 39%(-2), LAB 19%(+6), LDEM 31%(-8). This suggests that the Conservative vote in these seats is largely steady, with the Lib Dems losing support towards Labour. Most national polls have the Lib Dems down by much more than 10 points (between 10 and 15, depending on the pollster), so this suggests that while they are doing badly in marginals, it’s not as badly as national polls would suggest. If this pattern was repeated in Lib Dem held marginals, the Conservatives would make moderate gains from them.

The other questions in the poll show an electoral pretty much evenly divided. 47% think the country is going in the right direction, 49% the wrong direction; 48% think the cuts are too deep and too quick, 50% think they are about right or too shallow; 46% think things are difficult now, but will get better in the coming years, 51% think the economy will be no better in 3 or 4 years.