I agree with Assoc Professor Richard Herr's observation that neither major party can really take too much from the EMRS poll. See Richard reported in the Mercury and the Examiner today.

The headline result for the newspapers was that support for Labor has slid 9 percent since August 2006 while the Liberals have increased 4 percent. Today, support for the major parties is neck-and-neck which means the next election is up for grabs ... or is it?

As in anything to do with politics, it is not that simple.

The first thing to consider is the extraordinary high undecided response to the survey: 19 percent (which is much higher than Newspoll, for example, ever gets when it polls on the mainland). EMRS produce a table which excludes the undecideds, thus assuming undecided voters will vote the same way as decided voters (although that's not to imply EMRS's Tony Hocking believes this will happen):

If this outcome was reproduced at an election it would almost certainly produce a 10 Labor/10 Liberal/5 Green, or four Greens with one of the major parties returning 11 seats; either way a certain "hung" parliament. But I don't believe we can treat the undecideds this way.

The Green vote is highly unlikely to be anywhere near 22% at an election. Two reasons: the Greens always do better in opinion polls than they do at elections (and this is true Australia-wide); and the Greens vote has been fairly static for the past decade at about 16-17%, so a sudden 5% jump is not likely. The most the Greens can hope for is five seats (one in each electorate), although I give Braddon less than a 10% chance. They can't realistically win two seats in Denison, and certainly not two in Franklin.

Let's see what happened to undecided/other voters before the last election. The graph below is mine, using EMRS data that extends outside that covered in this current report.

EMRS opinion poll and 2006 election results

We can see that the situation prior to the 2006 election was very similar to that now: Labor scarcely over 30 percent, and undecided/other voters at 20 percent plus. At the 2006 election these non-major party voters went Labor at about a ratio of 3:1, enough to deliver a Labor majority government again.

So the pattern appears to be that between elections a large swag of Labor voters peel off Labor and siton the fence for a think. At the election these voters mainly go back to Labor, but the danger must be that this won't happen to the same extent this time: they could transfer to the Libs.

As Richard Herr points out, Labor will console itself that it has been in this position before mid-term; the undecided voters are really "soft" Labor voters registering a protest and will come back. This viewpoint has some currency: there generally is a drift back to the incumbent in modern Australian elections (the "narrowing"); and there is also is a tendency for undecided votes in opinion polls to have a protest element that returns as votes for the government when the voter has to vote "for real".

But there is a strong feeling about that the undecideds will be much harder for Labor to win back this time:

1. "Cost of government". It is a fact that each election gets harder for the incumbent to win. Last time Labor was seeking a third term, so now it is a fourth. Everything else being equal, this one will be harder.
2. Leader popularity. Premier Paul Lennon is not a popular leader, he has admitted that himself (we "don't understand him very well") in launching a "new direction" for the rest of this term. The Liberal leader Will Hodgman, by comparison (in fact, against any current Australian comparison), is a very popular opposition leader. This has to be a factor if it holds come March 2010. (See the EMRS report for leadership survey results.)

It is a piece of received political wisdom in Tasmania that the need for majority government is the most powerful vote determinate among the "average" voting population. This wisdom says that the undecided voters can be "scared" into voting for the party that can best win a majority. Part 2 of this posting (expect mid-May) will discuss how this might play out, and I will include a probability guide of the outcomes at the election.

Also, see the Poll Bludger for a discussion of the EMRS poll. You can post comments there if you wish.