Dr KEVIN BONHAM
Three months ago the electoral outlook for the Bartlett Labor government seemed reasonably promising, despite the difficulties it was facing. Its position had clearly recovered since the acrimonious departure of Paul Lennon, Green support was looking shaky and Will Hodgman’s little band of Liberals seemed solid enough but didn’t appear to be going anywhere in a hurry. Successive polls had suggested that Labor was in the same polling position as the Lennon government at the same point of its four-year electoral cycle, and that government had gone on to win extremely comfortably, although most commentators had given it no realistic chance of doing so.
The climb back from the end of the Lennon era towards retaining majority government was a tough one, but Labor was probably about halfway there. But the most recent EMRS poll ( HERE ) says the Government is back about where it was in May 2008 in the last days of Paul Lennon’s leadership, the only other time it has shown the Government “trailing” (see below) in the last decade. If (and it’s a big if) this single poll is accurate in what it says about changes in government support over time, then the Government has kicked itself right back to the bottom of the mountain. But instead of being almost two years out from the next election with a clean and fresh young leader, it now has just seven months to go with both the Premier’s ability to clean up the mess and his style of doing so under scrutiny. And if this really is the case, then the mountain could well be too steep.

As Peter Tucker points out, we should not get too excited about a single poll, however dramatic it is. This particular poll was taken at an exceedingly bad time for Labor, coming only days after the Liberals’ Vanessa Goodwin overwhelmingly won the Pembroke by-election, with Labor-linked candidates (albeit disaffected ones) outpolled and outlasted by the Greens and a local councillor. Most likely the Pembroke saga has amplified the poll swing against the government. But given the strength of the vote for Goodwin in that by-election, I see nothing unbelievable about the idea that at least a few points of Labor support have gone elsewhere within the last three months. We will need to see the next poll to see how much of this is actually real and lasting.

The Poll Cruncher gadget at Pollytics.com is a useful one for playing around with opinion poll margins of error, except that the current version gets hernias when asked to contemplate a double-digit negative swing. Nonetheless, depending on how you feed in the August EMRS results (with or without “leaning”, with or without “undecided”) it turns out that we can say with 97.5% confidence that at least half the difference between the May and August EMRS polls is real. (One in twenty polls are “rogue”, but they can be rogue in either direction.) We cannot say for sure whether that difference is caused by good and bad polling periods as opposed to by genuine shifts in voter sentiment over the last few months.

While the Liberals nominally “lead” by seven points on the raw figure and six points with “leaning” voters added, that must be treated with extreme caution. I’ve previously shown that in the leadup to the 2006 state election, nearly all of those claimed by EMRS to be “undecided” were actually soft Labor voters. If that is the case again, it may still be that Labor remains well ahead of the Liberals in public preference. However, they would not be ahead by enough to win an outright majority, and in any case, we cannot be sure that the “undecideds” of 2009 will still behave in the same way as the undecideds of 2005-6. It is difficult to extrapolate anything about the real standing of the parties relative to each other in these circumstances.

The poll is good news for the Greens, whose increase in polled support from 13 points to 17 is weakly significant. It is a little difficult to take the claimed 21% statewide Green support after excluding undecideds seriously; as I have noted before, virtually no undecideds actually end up voting Green. Also, EMRS tend to overestimate the Green vote, so if an election was held right now the Greens would be more likely to poll something like 17%. That, however, would suit them fine, especially if voters were fairly evenly split between the major parties.

Once again the electorate-by-electorate breakdown shows that EMRS have been sampling in a skewed fashion with more voters sampled in Bass (249) and fewer voters sampled in Lyons (142). I do not know why they are doing this and it does appear that they are not weighting the results to compensate for it, which if true is truly bizarre. That said, it rarely makes any great difference to the total. What difference (a point at most) it makes tends to overstate the Liberal position and understate Labor’s, since Lyons tends to be one of Labor’s strongest electorates, and not one of the best for the Liberals.

The electorate by electorate breakdowns are based on such small sample sizes that extrapolating seat-by-seat results based on them is extremely difficult. In most cases the seat-by-seat differences from the last poll are not even statistically significant, even though the differences in the two polls overall most certainly are. But one thing does stand out: by far the biggest drops in Labor’s polling have occurred in Denison and Franklin. The biggest Liberal rise is also in Denison, where there was a significant 12.5-point two-party swing between the May and August samples. The greater swings (albeit on tiny samples) in the south of the state are consistent with the Pembroke saga being the prime cause of loss of support for the Government, but they may also partly reflect the impact of budget cutbacks on largely Hobart-based public servants.

I am not going to attempt to combine previous electorate samples with these to get a larger sample size, because it is not appropriate to do that when there has been a statistically significant shift between one poll and another. However, it is worth running through the seat by seat breakthrough I gave in the previous poll to see what happens to it if a substantial swing to the Liberals and Greens is applied. Franklin still goes 2-2-1. Denison goes 2-2-1. Braddon starts getting interesting; assuming the Greens couldn’t really get 15 there then the flip from 3-2 to 2-3 (a possibility I generally rubbished while Labor’s numbers were much better) comes into play. 2-2-1 is also possible if the Greens can really lift. Bass stays 2-2-1 and Lyons goes 2-2-1. An election held on August 3-5 would therefore have probably returned 10-11 Labor, 10-11 Liberal and 4-5 Green. If nearly all of the undecided voters had gone one way or the other that assessment could change, but in any case a majority government would not be likely.

Polls like this aren’t very good for Labor’s chances. The vagaries of EMRS polling are not well understood by voters (although they do know these polls have been wrong in the past), and so if a poll shows Labor “trailing” by six points, it is rather difficult for them to run the argument that only they can win outright. In a piece written following the Pembroke by-election, Antony Green drew attention to the way Tasmanian voters swing between the major parties. Although huge swings from election to election are rare, Antony notes that between the 1992 and 2002 elections, roughly a quarter of the voters had jumped from Liberal to Labor, “a level of voter volatility unseen elsewhere in Australia.” This reversed a similar pattern that had developed between 1979 and 1992, except that in that case, the Liberals had not gone up so much because of the rise of the Greens. Tasmanians are thus unusually inclined to swing sharply between the major parties over the course of a few elections. Since our first experience of minority government reliant on Green support (1989-92), four of the subsequent five elections have returned majority governments. This is despite the difficulty of winning majority government in the Tasmanian system and in one case (2006) despite many reasons why that government could have lost its majority. But a government does not get to sit in majority just by endlessly scaring the electorate with the fear of a hung parliament – not if it annoys enough people to get itself thrown out anyway. This was amply demonstrated by the fate of the one-term Ray Groom majority government, which let power go to its head by voting itself a massive one-step pay rise that it failed to sell effectively, and by trying to kick the labour movement while it was down (reinvigorating said movement in the process).

Furthermore, once belief develops that a majority government really cannot retain that status, those begrudgingly voting for it have no reason to remain in the fold, which may result in a delayed swing that otherwise would have happened at the previous election. As such, if a majority government is clearly losing its majority then the swing can easily amplify, and we saw this in 1996 when voters walloped Groom with a 13-point loss of primary votes, most of which went to Labor. In 2006 Lennon managed to contain the swing to a few harmless points, but many more voters may have voted against him had they not seen the risk of a hung parliament as an issue. Labor needs a better poll result in October or the view that it cannot win majority government is likely to set in for good.

I do not like to make much of the “preferred premier” polling, since analysis at Pollytics and elsewhere has shown that it generally lags the real indicators of government support and is therefore relatively useless, whatever the Examiner thinks otherwise (probably subscriber-only by the time you read this.) It would be much better if somebody polled true approval ratings of the three leaders, which is a far more useful measure. Nonetheless, a striking difference between this poll and the May 2008 one is evident. The ratings of the Green and Liberal leaders are almost identical, but David Bartlett polls 30% compared to just 17% for Lennon in his final weeks of rule. Thus, many voters who had given up on Lennon without giving up on his government or embracing any alternative leader are still happy with Bartlett. Most likely, these were Labor voters who wanted Lennon replaced, and by shunning him as Premier in the May 2008 poll, they placed him in an untenable situation.

Looking at the electorate by electorate breakdowns of Preferred Premier and comparing it to the state of the parties, some really strange things become apparent, and they are interesting because although the sample sizes are tiny, the same people are being asked to compare leaders as are being asked to compare parties in each particular poll. Thus I have looked at the figure ((Bartlett-Hodgman)-(ALP-Liberal)) as an indicator of how Bartlett’s relative fortunes compare to his party’s.

This indicator has generally been on the slide at the rate of about four points per poll in recent polls, to the point where it can no longer be said that Bartlett’s popularity is racing ahead of his party’s. But while Bass, Braddon and Franklin have all shown a steady decline in the “Bartlett factor” over the last three polls, in Denison Bartlett took a hit compared to his party in the May poll, while in the August poll his party took a far bigger hit than he did. In Lyons there was no change in the “Bartlett factor” between February and May, but then in August Bartlett gave away 14 PP points of which Hodgman picked up 11, while his party only dropped 5 of which the Liberals picked up just 1. Tiny samples, but among those samples are some very odd voter attitudes.

Three months ago I assessed Labor’s chance of retaining majority government at 50-50 based on the then available evidence. Another three months of mess later that assessment (which was more optimistic than most on offer anyway) is no longer sustainable. Not only will Labor need to pull their socks up if they are to cut their 2010 losses to the virtually inevitable seat in Franklin, but I now doubt that they can retain majority government by their own efforts whatever they do. There would also need to be significant campaign errors by the Liberals or the Greens – as there certainly were by both in 2006.

Some final words about the final result in Pembroke. In one of the more implausible comments before the Pembroke by-election, it was asserted that the entry of Honey Bacon to the race would confuse the issue of who the real Labor candidate was and take the shine off a Liberal win since that win would only be on preferences. I immediately disagreed with this claim, on the grounds that a win, however narrow, over the famous Bacon name would only increase the prestige of any victory. In the end the margin over all opponents was so massive that waiting a week and a half for the result has only increased attention on the very decisive outcome and further highlighted Labor’s failure to attract a competitive candidate who would really fly the flag for the party (whether endorsed or as a token independent is irrelevant.)

Vanessa Goodwin won a normally Labor-leaning seat with a final three-candidate preferred vote of 51.1% compared to 27.7% for independent Richard James and 21.2% for the Greens’ Wendy Heatley. This is an extremely similar result to that obtained by Labor incumbent Allison Ritchie in 2007 (51.5% to James 31.2% and Smith (Green) 17.4%). It is also very similar to the result obtained by long-term Labor incumbent Michael Aird in the outrageously pro-Labor electorate of Derwent (51.6% to Branch 33.5% and Gunter (Green) 14.9%). I have already noted that Goodwin polled five points better than her party did in the 2006 state election, and that those votes most likely came from Labor. We now know that an additional thirteen points worth of habitual Labor voters gave their preference to the Liberals instead of giving it to the Greens or a well known, respected and politically moderate Independent. Considering that Goodwin would have also lost plenty of normally Liberal primary votes to Richard James and the other unaligned independents (and never got them all back since James came second) it is likely that something like 25 points worth of 2006 Labor voters in Pembroke are at least not allergic to voting Liberal when Labor is not contesting. If almost half Labor’s Pembroke support base is this open to voting for a Liberal, then Labor may find its support very soft around the edges indeed if the voters still perceive it as a mess come election time.