February 17, 2004

Campaign and election myths

I know I've been tardy with my final piece of analysis on the Queensland election, promised for last Friday, but in the meantime I thought I would post this piece by Warwick Powell.

It's an interesting analysis of some common beliefs about elections, although not one with which I agree 100%. I'd be interested in any feedback. Warwick Powell is a market research consultant and was a Labor marginal seats strategist and campaigner and adviser to the Goss Government. Contrary to gossip on Crikey! he assure me that he did not work for Merri Rose in Currumbin in the last election.

Campaign and election myths

By Warwick Powell

It’s time to dispel a couple of myths about political campaigns and electoral behaviour. These myths have either crippled effective campaigning (especially in the most recent State election by the Opposition) or have led to chronic misunderstanding of voting behaviour and the reasons for it.

Myth 1: voters prefer positive to negative advertising, and negative campaigns turn voters off.

There is a phenomenon in market research that is known as “learned responses” – that is, people repeat back to researchers what they have learned about the state of public opinion, thereby reinforcing the original point of view. This is exactly what happens with positive and negative advertising. When asked, voters these days will typically say they don’t like the negativity – but that’s largely because that’s what they think is the “right” thing to say "cause that’s what 'the published research and journalists says'”.

In reality, empirical research conducted in the USA – arguably the global leader in electioneering and behavioural research – shows that voters respond more to negative advertising than positive campaigns, provided that the negative messages can be substantiated and are in general terms “fair”, “credible” and not too “heavy handed”.

Strategically, the key is to ensure that one builds up one’s positives before one goes negative.

Political parties and candidates – in particular challengers – who fall for the line that they should stay “positive” for fear of alienating voters are doing themselves a chronic disservice.

The bottom line in this day and age, when voters generally hold politicians as a class in fairly low esteem, is that voters are more readily swayed “against” a politician than they are “for” a politician.

Just look what happened to Merri Rose in Currumbin as a case in point.

Myth 2: there is something called the “protest vote”.

The idea of a “protest vote” came to prominence when people went over the entrails of elections that didn’t fall the way commentators expected. So we have a situation where the defeats suffered by erstwhile popular politicians like Sally Anne Atkinson, Wayne Goss, Jeff Kennett and the near-defeat of Nick Greiner’s government in NSW is explained by reference to this thing called “the protest vote”.

Specifically, when people talk about the “protest vote” in these contexts, they imply that there is a difference between the voting actions of an individual and their “true” intentions about a government. In other words, a “protest” vote is when someone votes against the government but doesn’t want to vote the government out.

This is a nonsense, and is a self-serving myth promoted by those associated with losing campaigns who don’t want to face up to their underlying campaign and political failures and unpopularity. Invoking the “protest” vote idea is a way of saying it was “voters who made a mistake”.

How disingenuous is this? The fact is voters are making a considered decision when they mark the ballot paper – they vote with intent and, in their eyes, aren’t making a “mistake” or “protesting” with the end-result being “accidental” or “unintended”. Those that voted against Atkinson did so because they didn’t like her arrogance and perceived policy failures. Ditto for Goss and Kennett. They did not want these guys running the show, and gave them the proverbial a**e.

In Queensland, the Mundingburra by-election of 1996 should have put paid to the myth of the “protest” vote for good. This is for the reason that had the 1995 general election result (Goss’ near death experience) been a function of a misplaced “protest”, then voters when faced with the reality of the potential consequence of turfing out the Goss Government in the 1996 by-election wouldn’t have voted for the Liberal’s Frank Tanti. But they did.

According to the Oxford dictionary, “protest” is a “statement of dissent or disapproval”. This is the only sense in which the idea of a vote against a government being a “protest” actually means anything. When someone votes against a candidate or a government, it’s because they disapprove of it for one reason or another, and do not want the candidate or government elected. It’s not a mistake or an accident. It’s willful and entirely purposive.

Myth 3: the average swing is meaningful.

Almost all the coverage before and after the recent State election has focused on the state-wide swing against the Government (or swing to the opposition parties). There’s been some discussion about how best to "measure" the swing, especially given the "optional preferential voting" system that exists in Queensland.

By focusing on the average swing, analysts concoct a device called the "pendulum", from which they predict that for the government to lose a swing of X per cent is needed. Commentary glibly notes that this in fact is not quite the case, because the swing is not "uniform", but then goes on to discuss electoral outcomes in terms of what is actually the average "swing".

People who run campaigns know that practical realities don’t sit comfortably with the pendulum device. That is, it’s not the average swing but the distribution of the range of swings across electorates that determine electoral outcomes. For example, this is why One Nation was able to get some 11 seats in the 1998 election with around 21% of the vote across the State. One Nation’s vote was actually concentrated in pockets, which are best understood in spatial-demographic terms. The average swing to them (+21%) was entirely meaningless as a predictor.

Analysts should worry less about the average swing, and should focus their minds on the range of swings and the distribution of the vote. That way, they will be able to better predict campaign impacts and voting behaviour drivers.

When we look at things this way, we can dispel the last myth, namely that elections are "unwinnable" because of the number of seats involved and the size of the average swing required to cause these seats to change hands. It was this myth that sat behind the idea that for the coalition parties, the 2004 state election was really about positioning them for a serious tilt in 2007.

Most elections are winnable, particularly as the proportion of ‘rusted on’ voters declines; for challengers it’s a matter of figuring out ways of generating a sufficient range of swings against incumbent candidates in a large enough number of seats. Peter Beattie did it in 2001 and John Howard did it in 1996.

These campaign myths had real effects in the recent State campaign.

The Government was naturally worried about a “protest” in the Oxford dictionary sense, which could only be generated through an effective negative campaign from the Opposition. Labor effectively spooked the Opposition out of this kind of campaign, and consequently avoided the possibility of voters expressing disapproval. (I notice that Springborg has waited until after the election to challenge Premier Beattie on Freedom of Information laws!)

In any case, the coalition never thought they could win (“too many seats”, “too large an average swing needed” were the excuses), so their challenge was akin to being attacked by a wet lettuce leaf.

So that's Warwick's take. We might have to have an argument online about protest votes, or that there are no "unwinnable elections"!

Posted by Graham at February 17, 2004 04:28 PM


I don't know of any studies on this, which is a bit odd because we know that there are studies about how people buy retail products when they are in a store, which is used to decide how to merchandise.

We all tend to rely on the message being right and ramming it home as many ways as we can, including on the last day.

I'm not all that sure that manning polling booths counts for that much. Like you I've put in many hours handing material over in the nicest possible way, but the votes inside the booth never seem to reflect the massive effect my personality and style must have had!!! ;-)

I suspect that we put so much effort in on election day for reasons of our own personal satisfaction, and because it is the last part of the process where we have any influence.

In fact, I did once semi-scientifically study this in the constitutional referendum of 1988. There was a fair bit of dissent amongst Liberal Party members about the correct position on a number of the four propositions, and as a result, most people weren't prepared to hand out the party's How-to-Vote Card. I was the local Federal Electorate Committee Chairman and didn't have that luxury. So I took How-to-Vote cards around to the booths that weren't manned and left them there in plastic bags. Afterwards I did a comparison between those booths that were manned and those that weren't. The results were comparable, indicating that a presence on booths made very little difference in this instance.

Posted by: Graham Young at March 13, 2004 03:20 PM

As someone who has been involved at grass roots level in election campaigns for many years, what I really want to know is how voters decide on the day. When working at the booths I have found that many voters don't know which electorate they are in, so how do they decide? Is it by party of choice; is it the coreflute signs, if so, is it the colour, the photo of the candidate and if so whether it is female or male and the presentation casual or business like; or is it those dreaded How-to-Vote cards? Has anyone done any exit polling along those lines?

Posted by: Robyn at March 13, 2004 12:48 PM

Just an observation on a local campaign where the Labor candidate started campaigning some ten months out from election day and put in an enormous effort, she was a credible candidate who gave the National Party incumbent a run for her money, however, she concentrated on being POSITIVE. Maybe too positive?? Last time the incumbent only held on by approx. 63 votes, with the change in demographics, Labor should have taken it.

Posted by: Robyn at March 9, 2004 03:58 PM

I agree the protest vote idea has no empirical support and would instance The Entrance NSW by-election of 1991 just after the alleged 'protest vote' NSW election of 1991 as an example to go with Mundingburra and Frankston. I agree there are no unwinnable elections, if a government has a landslide victory that is a sign it is controlling the political agenda, and as a rule it usually takes some time for that control to evaporate. Beattie won in 2004 because voters approved of the government's performance in 2001-4 not in 1998-2001. From memory in Queensland 1995 and Victoria 1999 the opposition steadily gained ground in the campaign and the final result was within the error range of the polls (although the media ignored this). In the United States presidential landslides are fairly common but have little predictive value for the next election.
Electoral pendulums can be taken too seriously: 1) in most marginal electorates population shifts will exceed the margin; 2) the net swing in any election obscures larger aggregate movements of voters in both directions. I agree with Alex McConnell that by and large aggregate electoral outcomes in terms of seats reflect the aggregate movement of voters, although there are always a few electorates that defy the trend such as Noosa and Currumbin. Probably the swings match a normal distribution. The implication of this is that much of the electoral astrology about individual electorates that clutters the web is useless from the viewpoint of predicting an election outcome, even although it might be useful for predicting which electorates will deviate markedly from the aggregate swing. Some web commentary on individual electorates before the last Victorian election seemed to ignore the fact that the polls consistently predicted a Labor landslide.
Finally commentators have been a bit unfair to the Nationals in the Queensland wrap-up. By definition in a compulsory voting system to say that the percentage vote for a party (One Nation in 2004) has declined is the same as saying that the vote for at least one other party has increased (the Nationals in 2004). The Nationals won over One Nation voters and this was a positive achievement, those 2001 One Nation voters could have stuck with One Nation or gone to Labor (as a fair portion of 1998 One Nation voters did in 2001).

Posted by: Geoff Robinson at February 22, 2004 12:01 PM

A lot of the analysis of seats is too lightweight. I had a bit of a look at the demographic details on some of the seats in the Qld election and there were quite striking differences in age distribution and also in the amount of voter mobility into and out of electorates. Although it seems a reasonable simplifying assumption to say that electorates broadly maintain their demographic composiion between elections, I don't know if this is really true any more. Evven if they maintain their age and income demographics, however, the decline in rusted on supporters means that new voters who just happen to vote a particular way can be important.

I agree that the average swing is pretty meaningless. However, the two-party preferred vots is still a pretty good measure, subject to the difficulty in working it out in advance from polls where there is an optional preference system. If it can be worked out, it generally will give a pretty good indication as to what will happen, given that other factors occur more or less randomly, which will generally occur when there are a reasonably large number of electorates. My take on the Qld election is that the opinion polls two party preferred vote predicted a slightly smaller landslide to Labor than in 2001 and that is exactly what happened.

Posted by: Alex McConnell at February 17, 2004 10:29 PM

Thankyou for the insights.

Posted by: Ian Latto at February 17, 2004 06:09 PM
Graham Young
John Black
Mark Bahnisch
Michael Lee