Tomorrow is, as anyone who follows politics can hardly avoid being aware, the Eastleigh by-election. In many ways it is the most interesting and important by-election of the Parliament so far. We’ve had one proper Conservative-Labour marginal in the form of Corby, but given Labour’s strong lead in the national polls a Labour gain was almost a foregone conclusion. The most interesting sort of by-election in this Parliament was always going to be one in a Lib Dem-Conservative marginal.

There have been five polls of Eastleigh during the very short by-election campaign, two commissioned by Lord Ashcroft, two by Survation for the Mail on Sunday and one by Populus for the Times:

All the polls have shown the Liberal Democrats and Conservatives quite closely matched, three with leads for the Lib Dems, two with leads for the Conservatives. The only changes between the Ashcroft poll at the start of the campaign and the ones at the end of it is a further drop in the Labour vote and an increase in UKIP’s vote. From the polling you’d expect quite a close race, with the Lib Dems perhaps more likely to win (which, given the Liberal Democrats continuing very strong performance in local elections in Eastleigh since 2010, is what I’d have expected to see). A remaining unknown is what effect, if any, the Lord Rennard scandal has. The most recent poll by Lord Ashcroft was conducted after the Rennard story broke and shows no obvious impact at all, but clearly the story can continued to roll on since then and become increasingly about how Nick Clegg has handled the issue.

Come Friday we’ll know the result, and people will be clambering to declare what it “means”. Well, I’ll start with the same message I give after every by-election – the result won’t tell us anything about the national picture that we can’t get a much better handle on from national polling. By-elections are a creature that is more different from national elections than it is similar: they happen in only one constituency, in this case one where the Lib Dems are overwhelmingly dominant at a local level and have an unrivalled network of local deliveries and supporters; there is an intensity of campaigning and campaigning spending that dwarfs that in any general election and, most importantly, it makes no difference at all to the government of the country. It is just one MP, in one constituency, so however people vote the government won’t change, only their local MP will. By-elections are either different from national polling, in which case it is a result of the unusual circumstances of by-elections themselves and the particular circumstances of the seat, or they are very much in line with national polling, in which case they don’t tell us anything new.

Anyway, let’s assume the results are as the polls suggest – that the Liberal Democrats narrowly hold the seat over the Conservatives, UKIP do extremely well and Labour get squeezed down to fourth place. Journalists will write comment pieces concluding that the Lib Dems will do much better than the polls suggest as they’ll still be able to get tactical support from Labour, that UKIP pose a serious threat to the Conservatives and that UKIP voters DON’T seem willing to vote tactically for the Tories and, for the Conservative leaning amongst them, that the poor Labour performance shows that Ed Miliband’s “One Nation” mantra is just empty words.

All of these conclusions are nonsense.

Or at least, while some of them may very well be true, none of them will be things we can tell from Eastleigh. Firstly, if it votes in line with the polls Eastleigh doesn’t really show the Lib Dems withstanding the national swing in seats they hold. In the final Ashcroft poll they were down 14 points on their general election score, which is pretty much what the polls show is happening to their national support. Secondly, I wouldn’t conclude anything about tactical voting either Lab to LD or UKIP to Con – by-elections are very special cases, voters get an intense amount of literature and contact from the parties imploring them to vote tactically and send a message, and their vote won’t change who governs so in many ways people are free to vote without consequence. The argument about Labour’s one nation message is just point scoring – it is perfectly normal for a third party to be squeezed in a tight by-election and despite the exuberance of some Labour supporters at the start of the campaign it was bleeding bloody obvious from the beginning that Labour had no hope whatsoever in this seat.

While the by-election won’t actually tell us much, that definitely DOESN’T mean that its not important. On the contrary, I think whatever the result it will be extremely important in terms of party morale and the political narrative.

David Cameron’s backbenches are already extremely restless and for the Conservatives to win a majority at the next election the party need to win a substantial number of Liberal Democrat seats. If the Tories win Eastleigh David Cameron can reassure his MPs that they can take Lib Dems seats, if they fail to do so it risks increasing the unrest on the Tory backbenches and putting further pressure on Cameron and Osborne (especially if UKIP run them close – if UKIP beat the Conservatives then Tory backbenchers risk having a nervous breakdown).

The Lib Dems meanwhile definitely need a win to try and move the narrative on from the Rennard scandal. The blow of losing a seat on top of the ongoing crisis around the party’s response to Rennard could be extremely difficult.

The final possibility is a UKIP win. It is extremely unlikely – all three of the final polls had them in third place and 12 points behind the leading party. The general consensus is that they had the momentum in the election, and the Rennard story may have pushed more people away from the Liberal Democrats in the final days… but equally, a lot of people would already have voted by post anyway. It would suggest a problem with the by-election polls if UKIP did win. Nevertheless, were it to happen it would have a serious effect on politics, you’d expect a big boost in UKIP national support and the effect on Conservative party morale would probably be drastic.

So when the result comes in on Friday morning don’t look at what it tells us about public opinion – by-elections are by definition unusual – look at how it effects the political narrative.

Sky News have already released the monthly ComRes poll for the Independent here and show topline figures of CON 31%(-1), LAB 43%(+4), LDEM 8%(-2), UKIP 9%(-1). A significant increase for Labour and the largest Labour lead they’ve shown in their phone polls (albeit not significantly different from several 11 point leads they showed last year).

The poll was conducted between Friday and Sunday, so after the allegations about Lord Rennard and at least partially after the downgrade of Britain’s credit rating. Surely they are what’s caused the shift in the polls? Not so fast…

The Lib Dem score looks bad for the party, and it’s the lowest ComRes have shown the Lib Dems in their telephone polls. However, let’s not forget all our usual caution* – it is just one poll and the changes are within the margin of error. Wait and see if it is picked up in other polling.

The Labour increase meanwhile is less mysterious. When there is a movement in the polls it is natural to look at the last few days for an explanation for what might have caused it. Natural, but wrong, as in many cases the previous poll was a whole month ago and the cause of any change could have happened at any point. In this case the last ComRes telephone poll was just after David Cameron’s pledge to hold an EU referendum and showed Labour dropping by three points. As we saw in the YouGov daily polling, that narrowing of the polls from the EU pledge was very short lived indeed – so the most obvious explanation to the change in the poll is an unwinding of the referendum pledge effect, taking us back to the region of the 11 point lead ComRes showed in December.

And the downgrading of the credit rating? If it does have an effect it maybe too soon for it to show up in this poll anyway. Let’s wait for a couple more polls and see what happens.

(*unlike, it would appear, the front page of the Indy, most of Twitter, etc, etc. Sigh)

With under a week to go till the by-election tomorrow’s Times has a Populus poll of Eastleigh (£). Voting intention now stands at CON 28%, LAB 11%, LDEM 33%, UKIP 21%.

If one assumes that the Ashcroft poll earlier this month was conducted by Populus using the same methodology, then this suggests that the Conservatives and Labour have both gone significantly backwards during the by-election campaign, while UKIP have advanced into a solid third place.

I’m hoping we may have more Eastleigh polling over the coming days so I’ll update once it arrives or once we’ve seen the Populus tables.

UPDATE: Rick Nye has kindly sent me over the detailed tables for the Populus poll so here are a couple more things worth noting:

First, there is a very high level of don’t knows or refusals. Populus interviewed 1001 people, but a quarter said don’t know and a further 12% refused to give a voting intention. In a close race that high proportion of don’t knows obviously have the potential to change things.

Secondly, Populus don’t just ignore those don’t knows. Like ICM they assume that a proportion of them will end up voting for the party they did in 2010. This adjustment made no difference to the Lib Dem lead, but reduced the reported level of UKIP support (as they didn’t have many former 2010 voters to reallocate).

Without the reallocation of don’t knows the topline figures would have been CON 26%, LAB 12%, LD 31%, UKIP 25%. The reallocation of don’t knows has worked very well for ICM and Populus at general elections, improving their accuracy, but I remain uncertain about how well it works at by-elections. Without it the position between UKIP and the Conservatives is very tight.

Thirdly, everyone assumes that the UKIP support comes mainly from the Conservatives. It isn’t quite the case. 17% of 2010 Tory voters are saying they would vote UKIP, but so are 11% of 2010 Lib Dem voters and 10% of 2010 Lab voters.