Thursday, October 25, 2012

The Abbott Factor: Opposition Leader Ratings And Party Standing

Advance summary of this article:

1. The popularity of federal Opposition Leaders has been argued to be irrelevant because it does not seem to have a relationship to two-party polling in data since 1986.

2. However that appearance is misleading because (i) new Opposition Leaders taking over in weak party positions tend to have good personal ratings (ii) the patterns during Tony Abbott's tenure have been different to those before he was installed.

3. Excluding Abbott, there is a relationship between the ratings of established Opposition Leaders about whom the voters have clear views, and the Opposition's polling.

4. Although this relationship is gentle and usually drowned out by noise, it is capable of changing the result of a reasonably close election.

5. During Abbott's tenure, the relationship has reversed - the worse his ratings, the bigger the Opposition's lead, and this relationship has been strong until recently.  However, it is probably mainly the Government, not Abbott, driving this pattern.

6. The Opposition's lead appears to now be shrinking without Abbott's popularity greatly improving.

7. If that trend continues and the next election is close, Abbott's unpopularity could well cause the Coalition to lose.

8. Consistent with the above, no lastingly unpopular Opposition Leader has led their party to a federal election victory in Australia since 1950.

(NB 1. I will try to post advance summaries of long pieces like this one for the benefit of those who are either time- or detail- challenged.  I hope this will be appreciated by media in particular.)

(NB2.  Apologies in advance for the primitive appearance of the graphs. My Excel dates from the Clinton years.  I will find better graphing goodies someday.) 

Over the past two years I've been very interested in some unusual aspects of federal polling.  It has been uncommon, in the polling history of the last few decades, to have long periods in which both the Prime Minister and the Federal Leader of the Opposition were very unpopular.  It has also been rare to have periods in which the Federal Opposition is soaring in the polls but their Leader rates very poorly.  On my rolling sample of Newspoll data, what I call the Opposition's "disconnect" score reached its highest ever level in late July this year, at which time a minimum of 14.6% of voters were saying they would vote for the Opposition although they disapproved of Tony Abbott.

Many of my thoughts on these were covered in two of the longest pieces I ever wrote for TT:

When both leaders are loathed: a polling dynamics history (September 2011)
Don't write off Labor ... yet: A Brief History of Federal Polling (September 2012)

The picture has changed now, at least temporarily.  The PM is becoming almost popular, and the two-party preferred (2PP) gap is narrowing at a rate the Coalition might be quite alarmed by.  But Tony Abbott's ratings are still bad.

The question then is: do Opposition Leader ratings matter?  In the second of the above pieces I've pointed out that every federal Opposition Leadership that has recorded lastingly and substantially bad net satisfaction ratings (netsats) has ended in either election defeat or the Leader getting dumped.  If a leader as unpopular as Abbott ever won, it would have been in the first half of the 20th century, before polling existed.  (Unpopular Opposition Leaders have won at state level, but not that often, and that's irrelevant because federal politics massively influences state election results and can give unpopular Opposition Leaders an advantage.)

A recent piece by Possum Comitatus (Scott Steel), The Primary Dynamic, examines the relationship between Prime Minister and Government 2PP ratings in Newspoll.  It draws some startling conclusions about the relationship between PM net satisfaction and 2PP score in the lifetime of the Rudd/Gillard governments - variation in the PM's ratings has been driving (and it appears to be causal) very close to all the variation in 2PP.   This is not so clear for other polling firms and William Bowe in a comment on pollbludger (apologies for lack of link) raised the possibility that Newspoll secret herbs and spices (lots of pollsters have 'em, it's called "scaling") might be exaggerating the connection.

But when it comes to Opposition Leaders, Possum produced this and wrote "A drunk bloke shooting paintballs on a chart comes to mind with that particular graphic."  At a simple look, the graph says loud and clear that what matters is the PM's rating; voters keep the government if they think the PM is doing a good job, and turf it otherwise, and whether or not they like the leader of the alternative government has not a lot to do with it.

But it is actually not that simple.

The "paintball" graph is in fact the combined work of three paintball shooters.  An actual if rather weak netsat/2PP relationship for opposition leaders who are well established is being masked by the efforts of the paintball shooter in the (high netsat, bad 2PP) corner, representing relatively new Opposition Leaders, and the efforts of the paintball shooter in the (low netsat, high 2PP) corner.  And that last paintball shooter is none other than Tony Abbott, for whom the relationship between net satisfaction and party standing is completely different to that of other Opposition Leaders across the last three decades.

Here's a detailed breakdown of the problem.  My data source is a database of all federal Newspolls going back to 1986.

1. The More We See The Less We Like

New Opposition Leaders tend to start out with good ratings, probably because a proportion of their own party's supporters will tend to approve of their performance while supporters of the other party are likely to at least give them the benefit of the doubt in the grand old Strine tradition.  But familiarity breeds contempt and the following chart shows the net satisfaction ratings for Opposition Leaders prior to Abbott who have polled a given "uncommitted" rating:

That one's a bit messy because of the large number of data points and the crowding on the left-hand side is probably blurring the pattern a bit so here is what it looks like if we take the average netsat for each uncommitted value:

A new Opposition Leader usually starts somewhere on the right of this graph, with a high uncommitted rating, and moves to the left.  The trend is pretty flat while uncommitted exceeds 30 (though that section of the graph is based on very few actual readings) but for opposition leaders with uncommitted percentages of 30 or less it's not far from linear and it is very steep:

Netsat = 1.586 * Uncommitted - 30.35.

The suggestion is that as an opposition leader becomes better known, not only do the uncommitted voters tend to switch to disliking them (I suspect this is in almost all cases) but even some voters who liked them originally  switch to disliking them too. Thus many, but not all, opposition leaders record their highest approval ratings fairly early in their term while their uncommitted ratings are still high.
However, this relationship is so swamped by all the other noise that it explains only 10% of variation in Opposition Leader netsats. 

2. When Opposition Leaders Are New Their Parties Are Often Struggling

Here is the relationship between the proportion of voters who are uncommitted about the Opposition Leader and their party's lead (or deficit) in 2PP polling compared to the Government:

That one's a bit messy too so here's the same thing with just the average value for each Uncommitted percentage, and with a little bit of smoothing thrown in to stop those values on the far right, based on very limited data, from bouncing around so much:

I don't recommend putting a  line through that one. And not too much should be read into the up-kick at the far right since just two polls (immediately after Rudd and Downer were installed) have quite a lot of influence on those.  But what I especially want to draw attention to here - bearing in mind that these dots do not represent single data items but multiple readings averaged together - is the sharp jump just around Uncommitted = 20. 

Recently I did some work on bounce size for federal leadership changes (inspired by this post on Poliquant) and found that a leadership change for either government or opposition typically generates a bounce which peaks at about 3 points of 2PP (or 6 points netsat gap change) with this peak strongest on average about three to eight months after the change.  Some changes deliver only short-term benefit and as a result the average gap tends to fade, and it gets hard to track it beyond that because of leaderships ending.  Anyway, with the bounce finding in mind, I suspect the snake above should be chopped into five pieces:

Stage 1 (4% of Newspolls).  A new Opposition Leader has just taken over a party which in most cases was in a poor position to begin with (hence the change).  Naturally the 2PP is low.

Stage 2 (8% of Newspolls).  Includes a lot of bounce phase ratings.  The leader is probably still fairly new but the change in leadership has kicked in and the Opposition's standing has improved (on average to almost equality, though with enormous variation of course).

Stage 3 (13% of Newspolls).  The Opposition Leader on average has probably been there for a little while but the public still have not firmed in their opinions (for some leaders uncommitted ratings can still be over 20 after over a year, and for Mr Nice Guy Hayden they were for his whole career).  If the leader took over with the party in a struggling position, they've slipped back from the bounce, but the change has still had some effect.

Stage 4 (a narrow band on the averaged graph but 65% of Newspolls).  This is the serious combat zone.  The Opposition Leader is well established and not too unpopular and the party more likely than not has a slight mid-term lead.

Stage 5 (10% of Newspolls).  Normally, the zone of doom.  The voters have made up their minds and they don't like what they see.  The Opposition Leader is unpopular because their party is losing, or their party is losing because the Opposition Leader is unpopular.  Or both.  Time to start counting some heads.

So in summary of where we are so far (combining 1 and 2), relatively new (or perennially harmless) Opposition Leaders tend to be popular, but also tend to be leader at times when their parties are struggling.  For these reasons, the ratings of such leaders need to be discarded in order to look at the relationship between 2PP and Opposition Leader netsat in cases where the voters have a reasonably firm opinion about that leader.  

And here's that graph:

Again that's a complete mess because of the number of data points and the outliers, so smoothing the data points (as above) then averaging them in groups by netsats in bands of five, this turns up:

And there we have it, such as it is.  The high ratings on the right are mostly Rudd's but even if they're removed,  there was still something hiding in the paintball.  Excluding the Rudd stratosphere and assuming the rest (netsat <30) to be close enough to linear,

2PP = 50.51 + .0577 * Opposition Leader netsat, for well established leaders (uncommitted <21, netsat < 30)

It's only a medium-strength relationship that only explains about 10% of variation behind all the noise unrelated to what voters think of the Opposition Leader, and only docks an Opposition one point of 2PP for every 17 points by which their leader's netsat falls (cf. one point for every 7.6 points for a Government by Possum's figures, which has steepened to one point for every 5.8 points for Rudd/Gillard). 

And that was before Abbott.

There is only one graph that I need to post for Abbott, and I don't need to average or smooth out that one:

It's going the wrong way.  In Abbott's case, when he has polled terrible netsats his party has been way in front, but when he has polled even mildly positive netsats they have struggled.

Coalition 2PP = 49.77 - .205* Abbott netsat.  Quite a strong link with nearly 40% of variation explained.

I don't believe that this is strongly directly causative.  There is a tangled mess of relationships between Gillard's netsat, Abbott's netsat, and 2PP, and Gillard's netsat appears to be the strongest driver in it.  My current hypothesis is that, so far, when Gillard has been unpopular, this has brought out the most negative and aggressive side of Abbott, and of the mud he has thrown, some has stuck to the government and some has further damaged Abbott.

The damage inflicted on the normal relationship between 2PP and Opposition Leader netsat by the Abbott tenure has been immense.  For instance, the median Opposition Leader netsat for a pre-Abbott Newspoll with an Opposition 2PP of 54 or greater was +16.5. The Abbott tenure has destroyed that joint all the way to -1, which is little different from the long term median netsat for Opposition Leaders irrespective of 2PP.

Abbott also messes up the head of the uncommitted-vs-netsat snake (python or cobra?  hmmm...) by leading on 2PP when it says he should be well behind.  And the familiarity-breeds-contempt relationship is so strong in Abbott's case that for every voter who makes up their mind about him, his netsat goes down by a staggering 2.76 points (36% of variation explained by this alone!), leading him to record netsats on average 12 points lower than normal for such a well-known Opposition Leader.   (To be fair, if only for a sentence, much of this may just reflect a general disgust with the state of politics or even be led by disapproval of Gillard and Labor.)

What we are seeing in recent weeks is something that the Coalition should be quite afraid of if it lasts much longer.  As the 2PP gap narrows (and it's narrowing fast), Abbott's ratings have to this point not much improved.  The fear for them must be that the massive blowouts in 2PP seen over this government's term are in large part the result of government strategy (eg getting the carbon tax implemented well before the election) and earlier mistakes that the government has remedied.  So if this is a close election, as most of them are, a more normal Opposition Leader netsat/2PP relationship resumes, and Abbott's ratings don't snap back, then Abbott will just be one or two 2PP points of deadweight. And if he does something really silly, as happens every, er, week or two at the moment, it could be more.

It's notable that by the standards of the relationship existing before Abbott, the Coalition would have been predicted to have won the 1990 election had Peacock had a zero netsat.  They might not have won the 1987 election had Howard had one, but they should have won the 2PP and won more seats.  And, indeed, it's arguable that even Abbott's mild unpopularity before the 2010 election could have cost the Coalition just one seat, compared to had he had a netsat of, say, 5 or 10. 

I'll be keeping an eye on this one ...

No comments:

Post a Comment

The comment system is unreliable. If you cannot submit comments you can email me a comment (via email link in profile) - email must be entitled: Comment for publication, followed by the name of the article you wish to comment on. Comments are accepted in full or not at all. If you submit a comment which is not accepted within a few days you can also email me and I will check if it has been received.