Thursday, October 11, 2007

Playing with election predictors

I don’t know if it is just an Australian thing, but the only place I ever find people trying to fathom the arcane electoral prediction arts is downunder. Regulars will know I have a passion for divining election prediction formulas.

We have a US wingnut plonker Frank Luntz here through our faux election campaign working his magic. But young Frank is more into massaging numbers than predicting. He would rather push a poll than analyse polls.

Thus far I’m happy with my sledge hammer, household economic predictor- at least in its limited reach. In country such as Australia which has, essentially, national taxation collection, the theory works for national elections.

The household economic method is a blunt instrument which does not allow for many variables outside a simple two party/federal dynamic. It doesn’t work with states subject to federal disbursements or to a growing move to federal alternatives such as minor parties or independents.

So another Aussie has just posted Inertia, proximity and alterity , another look at predictor models. I don’t disagree with the writer, but I can’t post a comment on the site and I find the choice of words problematic.

Inertia I might have used traction; alterity might have better been alternatives. But given that the basics are important.

  • Oppositions do not win; governments lose
  • Australians [or voters generally perhaps] do not tend to change government in the absence of a recession, or policy failures
  • Inexperienced Opposition Leaders usually lose. Recession-free and debacle-free Prime Ministers usually win
  • Governments lose when the economy turns sour, or when ministers are found to be corrupt, grossly incompetent or excessively arrogant
  • Once elected, governments tend to serve at least two and often three or more terms
  • Governments which win their first term by the barest majority, and provided they have not stuffed up, often go on to a win a more substantial majority in their second term

I would add one more indicator to this list:

Party infighting is a clear sign of an impending loss for a governing party, but it also signals continued loss for a minority party. I’m not sure whether infighting is a product of failure to connect with the electorate or whether the electorate punishes parties which can’t get their act together. I suspect the former.

Now the economic failures are relative, as we saw in the US Mid Terms 2006. The Republicans could clearly claim a booming economy at the top end, but the claim did not extend through to household budgets.

In fact reports suggest that the Republicans were caught out during campaigning because they failed to identify the wider economic implications. I suspect the Democrats were as bad in this, but the electorate was not holding them responsible. We go back to the first point, governments lose.

When Paul Martin led the Canadian Liberals to elections in 2004 and 2006 the economy was in good shape, though reality varies again by strata. On international criteria the country was powering along, but household economics might not have been so rosy.

But Martin had inherited a tired and corrupt government, and nothing short of comfortable household economics could have saved him. The 2004 election threw his government into minority and the 2006 outing delivered the axe.

It was a sort of mixture of dynamics; the economics was simply not comfortable enough to override the legacy of corruption the Liberals had accrued over the years of Jean Chretien and Martin failed to convert his economic experience into a valid leadership position. Party infighting developed in 2004 and increased into 2006.

John Howard’s Liberals, in Australia, are talking up dubious economic success. To be sure, the economy is booming, but fiscal restraints and egos limit the benefits through the wider community. Howard has always been a wedge politician, so it’s not surprising that he is being caught in his own trap.

The impending disaster for his government has been closing in for over a year now, with polling consistently running against them. Howard’s mob has failed to gain any traction at all, despite the ability to wave wads of cash about.

They look tired and desperate; there is increasing dissention within the party at all levels. Leadership challenges have died off for now, but voters have taken note that none of the challengers were good enough, and one suspects, they’d rather punish Howard than a surrogate.

I could go on with examples, but they don’t really help with applying predictors to other dynamics. Lets face it, the US primaries institutionalise party infighting, so create a sort of level playing field.

But I will be visiting some of the variants in future to try and get a handle on application of these and other predictors. The whole topic still fascinates me.

2 comments:

D.K. Raed said...

Do you have Halloween or something similar down under, where the kiddies all don scary outfits & beg for candy? If so, you may be able to accurately predict your next election by noting which candidate's rubber fright masks are the biggest sellers. I swear, I'm not making this up! Some company here in the US ran statistics to prove this was as accurate a predictor as any poll. I think they looked back as far as Nixon. Everyone wanted a scary Nixon mask. Amazing!

Oh & my condolences if you have Frank Luntz meddling in any part of your elections. Him & his three-word test audiences have totally destroyed american's attn span for anything longer than a sound bite.

Cartledge said...

DK, thanks. Its wonderful to have something to bring a smile to the dial. I'm well versed in your culture, and have recently returned - reluctantly - from a stint in Canada.
The Halloween thing doesn't really work here because of the pervert factor, but I expect John Howard masks would kill it if we did celebrate.
I believe there is also a Washington sports team (not my forte)whose wins and losses are used to predict presidentials.
I love the 'off the wall' methods, but I'm probably too steeped in the evidence based methods.
I would really love to compile and publish the lot one day; chicken gizzards through to polling predictors.