With these questions in mind, much closer attention needs to be paid to the regular opinion polls that ask voters about what they think of the performance of the prime minister and the leader of the opposition. These polls provide newspapers with a lot of prime content and big headlines but not much else, it seems, in terms of generating serious thinking about these issues. In particular there needs to be much deeper analysis by the media of the contradictions inherent in the popular responses to leadership opinion polls.

The broad outline of community views on our leaders that is provided by the two big polls, Newspoll and the Herald/Neilsen Poll, are quite well known.

The last Herald/Neilsen Poll showed that Tony Abbott was preferred as Prime Minister over Julia Gillard by 50 per cent to 42 per cent with 8 per cent uncommitted. But there are two caveats. First neither leader is popular, though Gillard is especially unpopular. Only 44 per cent approve of Abbott's performance (52 per cent disapprove) and only 35 per cent approve of Gillard's performance (60 per cent disapprove). Second, with the Coalition leading Labor in this poll by 58 per cent to 42 per cent in two-party preferred terms Abbott is not as popular as his party and, it would appear, neither is Gillard. The pattern of the latest Newspoll is broadly similar, though it reports an enormous 24 per cent uncommitted on the question of better PM.

Given that pattern the Herald/Neilsen Poll goes on to ask voters about their preferred Labor leader. Here's where it gets interesting. The Sydney Morning Herald report begins with the proposition that Kevin Rudd continues to lead Julia Gillard as preferred Labor leader by 62 per cent to 30 per cent. But that figure is distorted by the overwhelming pro-Rudd preference of L/NP voters, the much larger group in the electorate at the moment, who prefer Rudd to Gillard by a massive 70 per cent to 17 per cent. Labor voters actually prefer Gillard to Rudd by 52 per cent to 46 per cent. The Sydney Morning Herald front page report notes this fact but doesn't make much of it. There are a number of implications. One is that despite being at such a low ebb Gillard retains the support of Labor voters, which, in one important sense, is what should matter, but that Rudd is still remarkably popular. But why are Coalition voters so anti-Gillard?

One possibility is that Rudd is out of sight out of mind. Another is that they detect particularly unlikeable qualities in Gillard. Yet another is that they are just taking Abbott's lead. In his budget reply speech he called for Labor to replace Gillard. Is this really what Abbott wants Labor to do? He has also called for an early election which, presumably, is something in Gillard's hands unless he wants Labor to change leaders but then have an immediate election. That is a big ask.

When asked furthermore should Labor change leaders, The Sydney Morning Herald gives prominence to the overall figure which is that 50 per cent of the electorate want to change leaders and 45 per cent say stay with Gillard. But again this is a misleading figure because it is distorted by the opinion of Coalition voters. They want Labor to change leaders by a margin of 65 per cent to 30 per cent. Yet Labor voters say stay with Gillard by 68 per cent to 29 per cent. Incidentally Greens voters say stay with Gillard by about the same margin (66 per cent to 29 per cent).

Labor voters make an interesting study in their loyalty and preference for Gillard, but by a big margin Coalition voters are much more interesting. What do they really want and why?

One reading of these polls is that Labor would do much better under Rudd. That is the case if Rudd's greater personal popularity with newly disaffected Labor voters and long-term Coalition voters will bring both groups across to the Labor fold.

Yet there is a contradiction. Some Labor voters admit this possibility that Rudd as leader might be good for party fortunes, but still want to stay with Gillard. Presumably their motivation is either loyalty or a belief that it is now too late to change or that it is inappropriate to change leaders at any time. This line of thinking is explicable.

Less explicable is the reasoning of Coalition voters. They appear to be caught between two stools. Surely they would want Gillard to stay put as PM on the grounds that her unpopularity makes eventual Coalition victory more likely. But instead they say that they want Labor to change leaders. Surely in electoral terms this is strange thinking. If they like Rudd more than Gillard then isn't there a danger for their side of politics that if Labor were to make him leader then it would do better at the next election? Perhaps Labor would do so much better that the Coalition would even be in danger of losing that election. That surely is not what Coalition voters want. But perhaps they dislike Gillard that much.

All this is a version of the old conundrum that any Opposition should secretly wish for an unpopular government leader to keep their job until the next election while publicly trying to drag them down as soon as possible by exposing their weaknesses. In that sense it is not an unusual situation. But neither Coalition voters nor the Leader of the Opposition appear to have thought this through at the moment. Nor have the media outlets that report on the opinion polls.

John Warhurst is an emeritus professor of political science at the Australian National University. John.Warhurst@anu.edu.au