Tuesday, November 17, 2015

How Bad Is Bryan Green's Rating?

Last week I reported on the EMRS poll which has shown a probably Turnbull-led resurgence for Tasmania's Hodgman Liberal state government.  One figure has dominated discussion of the poll in the subsequent week:


19% is the preferred premier rating of Labor Opposition Leader Bryan Green (compared with Will Hodgman's 56).  I often rail here against the media overuse of preferred-leader stats to spin a yarn, and against polls that poll preferred-leader scores but don't also poll and release approval ratings.  I even have a piece here declaring such scores to be rubbish.  They're not totally meaningless, but they're messy indicators that are often biased to incumbents, they lag behind changes in approval rating, and they don't have a very good predictive record.

Part of the problem is that a preferred-leader score is a comparative indicator, so it's impossible to discuss what it says about one leader without thinking about what it says about another.  Does a big lead for an incumbent Premier say that voters really like the Premier and don't mind the Opposition Leader, or does it say that voters mildly like the Premier and can't stand his opponent?  EMRS have been polling the answers to these questions, but unfortunately they haven't been releasing the results.

As Matt Smith observes (in a notable piece that suggests Labor are disheartened and just going through the motions) numbers like this can spell a lot of trouble for a leader, and can ignite leadership speculation.  We shouldn't overstate the "trouble" angle; some state leaders have dragged on for years with miserable polling, but there have also been cases interstate where just one poll like this has been game over.

Anyway, there's been some leadership-speculation this last week, but it's only very long-term stuff.  It's agreed that nearly two years on from a heavy defeat there is still nobody else "ready", so why discard an experienced leader and trash one of Labor's possible future stars? Especially when we still have more than two years until the next election. There's plenty of evidence that if a party is in a winnable position, and has a competitive leader in waiting, then nine to twelve months is ample time to successfully install a new state leader.

It's pretty hard to see Labor doing much worse in seat terms in 2018, so there aren't any panicking MHAs who fear that if party polling can't improve that their seat will be lost.  So this sort of polling doesn't alone justify a leadership change now, and may well not do so next year either.

Why Green's 19 Is Not Like Lennon's 17

As messy as these questions are, I think it's time to explore what this sort of rating for an established Opposition Leader might actually mean.  What might it say about what voters think about the leader, and what might it say about his prospects? The main reason for doing this is that Bryan Green's 19% rating has been widely compared to the infamous 17% from former ALP Premier Paul Lennon's final poll before he resigned.  The direct reason Lennon's stocks had sunk so low was leadership speculation that David Bartlett was going to replace him, leading Labor supporters who preferred Bartlett to pick the none-of-the-above option.

However, any comparison between Green's 19% and Lennon's 17% is futile.  Firstly Green is an Opposition Leader, whereas Lennon was a Premier, so on the surface it might be thought that Lennon's figure was a great deal worse.  Indeed, if we compare these to the all-time federal and state worsts for Opposition Leaders and Prime Ministers/Premiers in Newspoll, Green's is twelve points above the worst for an Opposition Leader while Lennon's is ten points below the worst for a Premier or PM.  (For those wondering, the worst OL score was 7% for Brendan Nelson during the Rudd honeymoon, while the worst PM/Premier score was 27% for Paul Keating following the 1993 horror budget, and also for WA Premier Richard Court just after he had beaten a Labor government led by the extremely popular Carmen Lawrence.)

That's not the full story either, because EMRS uses a different scale now to then - and this is a recent change, which makes it hard to compare current preferred Premier ratings to EMRS's history.  Currently the options it gives for Preferred Premier are Will Hodgman, Bryan Green or "Unsure/Neither".  However, at the time of the Lennon shocker, the readout included not only Lennon but also Hodgman (the then Opposition Leader) and the then Greens Leader Peg Putt.  The figures were 39% for Hodgman, 17% for Lennon, 14% for Putt and 30 for "None of the above"/uncertain.  Greens voters disliked Lennon for his forestry connections, so it might seem likely that if the poll had been just between Lennon and Hodgman, Hodgman would have been preferred by, say, 44 to 22.  However, even that is not straightforward.  For whatever reason, including the Greens in a preferred-leader readout seems to blunt the incumbent's advantage.

So we have a sort of case that Lennon's standing at the time of his departure could have been historically terrible, but it's not clearcut (and see also the Caveats section below).  It's easier to see how Green scrubs up compared to other Opposition Leaders who had been in the role for similar periods and polled similar ratings.

Green's Predecessors

I thought I'd have a look at the fates of other Opposition Leaders at state and federal level who have polled similarly poor preferred-leader ratings (anything below 20) in Newspoll after 18 months or more in their position.  For these purposes I've ignored all three-way contests, which removes most of the few previous Tasmanian Newspolls from the mix. The comparison comes with some important caveats, for which see the next section.

The following state Opposition Leaders qualified for inclusion: Kerry Chikarovski (Lib, NSW), John Robertson (ALP, NSW), John Brumby (ALP, Vic), Denis Napthine (Lib, Vic), Robert Doyle (Lib, Vic), Rob Borbidge (Nat, Qld - in two separate terms), Mike Horan (Nat, Qld), Lawrence Springborg (Nat, Qld - first term), John Olsen (Lib, SA), Rob Kerin (Lib, SA), Bill Hassell (Lib, WA), Barry McKinnon (Lib, WA), Colin Barnett (Lib, WA), Paul Omedei (Lib, WA), Eric Ripper (ALP, WA).

The following federal Opposition Leaders qualified: Simon Crean (ALP), Kim Beazley (ALP - second time), Bill Shorten (ALP).  I haven't kept old Morgan data but obviously John Howard in his first term would also make the grade (Newspoll only polled these ratings at election time back then).

In the time Newspoll has been polling, I found 20 Opposition Leaders with completed careers with at least one Newspoll after 18+ months who did not poll a preferred-leader score below 20, and the above-mentioned 18 Opposition Leaders with completed careers (counting Borbidge twice but not Shorten) who did poll at least one such rating.

The federal Opposition Leaders who polled bad preferred-leader ratings also had (or in Shorten's case, has) very poor net satisfaction ratings at the time.  But for the state Opposition Leaders who did so, this wasn't always the case.  They polled net satisfaction ratings at the time ranging from -27 to +14, with an average personal worst during the time of -13 and an average personal best during the time of -4.  They polled satisfaction ratings ranging from 21 (Chikarovski) to 45 (Kerin), with an average range of 31 to 36.  Several of them polled much worse preferred leader scores than Bryan Green, with Napthine's worst being 8% and best only 15%; on average they got down to 14%.

So, even if an established Opposition Leader is polling a pretty lousy preferred-leader rating (which is quite a common event), that isn't reliable evidence that the voters have a very negative view of them.  The average personal ratings for such a leader are something like 33% approval and 42% disapproval.  The preferred-leader rating is likely to be more about the uncompetitiveness of the Opposition and/or the Premier's popularity (two things that strongly tend to go hand in hand.)

However, the difference in fates between established Opposition Leaders who polled sub-20 preferred-leader ratings and those who did not, is very striking.  Of the 20 with completed careers who escaped this blemish, 11 led their parties from Opposition into Government during the same tenure as Opposition Leader, though some took more than one election to do it.  (Those seem like quite good odds for Mark McGowan or Steven Marshall, for example.)

Of the 18 with completed careers who did poll such low preferred-leader ratings, 12 did not even make it to the next election before being removed - though some became Premiers of their state much later.  (Some had already lost an election before polling such bad ratings and being replaced.)   Of the six who were allowed to face the people, only one didn't clearly lose in that attempt at the job.  The exception that proves the rule was Rob Borbidge (first time), who survived preferred-leader ratings as low as 14% as an established Opposition Leader to lead his party to within a whiff of victory in the 1995 Queensland election.  After one Labor seat win was annulled and Labor lost the by-election, Borbidge became Premier the next year.

So that's one win of a sort, five losses and twelve booted before the next election, from a sample of eighteen.  Grim.  And grim in theory for Bill Shorten too, except his current bad ratings might be more directly affected by the Turnbull honeymoon, so we should see if he's still getting them in a few months.

If we are to take Newspoll as a guide, this EMRS result is saying that Bryan Green will probably not even make the next election as leader and that if he does make it he will very probably lose.  While this is a widely-held view anyway, there are still some reasons for caution:


Here are a number of reasons why comparing Newspoll with EMRS ratings is difficult.

1. They don't ask quite the same question.  

Newspoll asks "Who do you think would make the better premier?"  EMRS asks "Who would you prefer to be the Premier of Tasmania?"  It's possible to prefer that A be Premier (for instance if A is the leader of the party you support) while also thinking B would actually be better at the job.  So it's not completely clear results from these two questions mean the same thing.  That said, when comparing results based on "better" vs "preferred" wordings across different pollsters, I've never seen major differences.

2. They are taken over different time scales.

Newspoll state polls are normally taken over a period of two to three months.  EMRS polls, however, are taken over a period of several days.  If a leader's standing goes up or down briefly for whatever news-cycle reason and EMRS happens to be polling at that time, then EMRS' results could be influenced, while the impact on Newspoll's would be lower.  A bad leader rating in a state Newspoll is more reliable as a sign that a leader was struggling for an extended period.

3. EMRS skews against Labor.

At elections EMRS polling tends to underestimate Labor by a few points and overestimate the Greens by a similar amount.  This might be caused by sampling strategy, but I think the real cause is that some voters who will actually vote for Labor claim to be uncertain, while some voters who are actually uncertain say they'll vote Green (or Independent) as a kind of glorified "dunno".  If that's all there is to it, then that's fine, but if it's a sampling method issue then it's possible EMRS ratings could be harsher on Labor leaders than leaders of other parties.

4. Landline polling is getting harder.

Most of the Newspoll state polls I'm making comparisons with date from the golden ages of landline polling, when almost everybody had a landline and a lot of people actually answered it.  This is no longer the case and the old Newspoll was starting to get some curious results in its final year or so before being replaced by the new Galaxy-run Newspoll, which is partly conducted online.  Although the old Newspoll got away with random-sampling by landline only at the 2013 federal election (with excellent results), there's a feeling that even in two years things have become much harder for purely landline methods.

I'm not even sure whether EMRS are still attempting a purely random sample of the whole Tasmanian landline base, given the frequency with which some numbers are called.  In any case, there is more room for doubt about this sort of result than about a similar result in a poll, say, ten years ago.

5. The circumstances

Lastly, federal politics affects state politics.  The replacement of a really unpopular PM with a so far genuinely popular one mid-term is actually an unprecedented polling event.  In some respects, voters are reacting as if what we have federally is not just a new PM but a new government.  Some of this may well be rubbing off in responses to the EMRS poll.  That said, plenty of the leaders in the Newspoll sample would have polled their bad ratings while a federal government was polling through the roof for whatever reason, and it didn't seem to help them.

So all up, while there's an argument that Green's recent ratings spell big trouble for his leadership down the track, there are also reasons not to be too sure of that.  If such ratings continue in further polls, then some of the latter reasons will disappear.


  1. Of those times where the poorly rated OL was rolled before the next election, how well have their parties gone at the next election?

    1. Good question. Five won and seven lost. Of the five wins, in two cases the replacement OL was also rolled before an OL was found who won.

      I suppose what this is saying is that oppositions are more likely to keep underperforming opposition leaders if they think they have no chance of winning anyway, rather than that rolling the leader automatically creates quite a good chance of success.