18
We're going to get a lot of polls, I'm sure, during the campaign so I just want to give some hot tips for reading the polls to people who haven't followed a GE that closely before.
- Changes within 3 points or so aren't statistically significant on their own. The thing to do is to look for overall trends. If this company does another poll tomorrow showing Labour on 42% or 47%, that really wouldn't mean anything much.
- This also means that, if a poll the day of the election looks like this, then any actual election result with Labour between 42% and 48% would mean that hypothetical poll was accurate - or no more inaccurate than expected. Obviously, that is a very wide margin, which should tell you something about the predictive power of individual polls this far out from election day!
- Don't look for whatever headline caught your eye lately as the 'cause' of the any shift in the polls. Almost nothing makes much of a difference to polling and it's almost certain that large chunks of the population missed whatever you thought was important. Things like partygate and Trussonomics really did move the dial: it's stories of that kind of magnitude that have a real impact.
- A corollary of the above is that almost nothing parties do during election campaigns makes a difference - 2017 was very unusual in that respect. Generally, voters have made their minds up already.
- You can't straightforwardly compare polls by different companies. If some other company releases a poll tomorrow with Labour on 41%, that does not mean Labour's lead has in any sense fallen over that 24-hour period. Again, you need to look at overall trends to have any understanding of what's going on.
- Relatedly, always look at the dates of the fieldwork and the dates of the changes (which OP has very rightly posted here). Some companies publish polls more often than others. Sometimes you'll see a poll with a massive change, but it turns out to be comparing with the last election, while many of the very frequent polling companies are comparing with last week.
- On a similar note, look at where the polling was done. Polls of, e.g., just London, which some companies do, tend to show massive Labour leads, which people sometimes get very excited about because they wrongly think they're national polls.