What causes poll movements

The Guardian's report on their poll this month has made the assumption that a drop in Conservative support is due to NHS policy, various other commentators have this morning jumped to the same conclusion.

The truth is, as ever, rather more complex than that. Firstly we should set aside the ICM poll apparently showing a 4 point drop in Tory support as the NHS row continues. This is a classic case of making the outlier the story, the media's constant failing in reporting polls - the underlying trend in polling is ignored while the rogue poll (and the reversion to the mean afterwards) gets a headline. Looking at the broad range of polls the Conservatives have not lost four points of support: they have gone from a position in December when, after the European veto, they appeared to be slightly ahead of Labour in the polls to a position this month when they appear to be a point behind or so, a drop of one or two points.

Many commentators today have also fallen victim to a fallacy I often see in the comments here - ascribing whatever movement happens in the poll to whatever subject they personally feel strongly about. Hence the Guardian and many bloggers feel strongly about the NHS, there has been a slight drift downwards in Tory support, therefore the former is probably the cause of the latter (it's also a sort of availability bias - the poll contained questions about the NHS, they are bad, therefore that's the cause. What if the other questions in the poll had been about, say, crime?)

However, this ignores other possible explanations, which could actually be better evidenced. Here are a couple, though I certainly wouldn't claim these are exhaustive - there are no doubt other possibilities, including those that are less well served with tracker polls.

The first hypothesis is the unwinding of the European veto effect. You will remember that in November Labour had a lead of four or five points. The Conservatives then pulled level after David Cameron's veto at the European summit. We should be expecting the effect of the veto to gradually unwind and, indeed, there is evidence to suggest that is what we are seeing.

The proportion of people thinking that Europe was an important issue facing the country peaked at 38% just after the veto, but has since faded away again, and is now back down to 23%. Similarly we saw some sharp increases in perceptions of David Cameron on the back of his veto. The percentage of people thinking Cameron "sticks to what he believes in" rose from 26% to 39%, decisive went from 20% to 29%, good in a crisis went from 13% to 18%. Since then most of these figures have fallen back down a bit ("stick to what he believes in" back to 31%, decisive back to 24%, good in crisis back to 18%).

One straightforward explanation therefore is just the fading of the leadership boost that Cameron received from the veto.

A second hypothesis is less negative coverage of Ed Miliband. In January the media narrative was dominated by criticism of Ed Miliband's leadership, and the media impression was increasingly that he was a no hoper. In February that has faded a bit, the attacks have mostly stopped coming, he's had a few good PMQs under his belt and the media narrative has largely moved on (to a great degree to the NHS!). This is reflected in Miliband's own ratings. In January he hit record lows in his approval ratings, reaching a nadir of minus 53 points. In February he has recovered a but to the low minus 40s. His ratings are still significantly worse than they were last year... but he has risen off the canvas.

A third hypothesis is the NHS. The recent coverage has certainly pushed the NHS up the agenda - going back to the YouGov issues tracker, the proportion of people naming the NHS as an important issue is up to 32% from 22% a month ago. We also have lots of questions that have shown the government's policy is unpopular and that people don't trust the Conservatives on the NHS, but these don't necessarily indicate change - it could've been unpopular all along.

The ICM poll yesterday shows that people trust the Conservatives less on the NHS now than in 2006, which is an interesting finding in itself, but it doesn't follow that the drop in trust has come in the last few weeks or is a result of the current policy. If we look at YouGov's regular tracker on which party people trust the most on the NHS there is a distinct lack of any recent drop in the proportion of people who prefer the Conservatives on the NHS.

In fact, the Conservative trend in the last year is pretty flat - what drop there was happened in late 2010 as the government's honeymoon faded. What is also worth noting in this graph is that the Conservative detoxification on the NHS is something of a myth - it never really worked very well, and Conservative leads on the NHS were small and transitory.

This, I suspect, is one of the reasons the NHS hasn't damaged the Conservatives that much - most people never trusted them on it in the first place. The other reason is that most people have no clue about the reforms and what they are or are not likely to do - take this recent YouGov poll for the Sunday Times which, rather than tell people about the reforms and ask about them, just asked people to say if they supported or opposed them based on what they'd seen or heard about them. 48% of people were opposed... but over a third (34%) of people said don't know, a comparatively high figure for any question.

That is not to say that the NHS reforms are not politically dangerous for the government - I think Tim Montgomerie's point that any future problems with the NHS (and there will always be some problems with the NHS, even if there were not budget squeezes and reorganisation) will be blamed on the policy is well made. Equally, just because it hasn't made a big impact so far doesn't mean it couldn't do so in the future.

However the claim that it is already doing significant damage the Conservatives in the polls is weak. The drop in Conservative support in the polls is small and there are alternative and perhaps better evidenced explanations for it.