When do voters make their decision?

In a post at his very useful Oz Politics Blog during the week Bryan Palmer said:

We know from past election studies that roughly half the electorate reports that it decided how to vote during the election campaign.

The source of this statistic is probably the Newspoll conducted after each election that asks voters which party they voted for and then follows up with ‘when did you yourself finally decide to vote for …. party?’ In 2004, 30% of voters gave various times in the last week, and another 19% in the last month (ie, during the campaign). That makes 49%. Similar numbers were recorded in the 2001 election (46%), the 1998 election (50%) and even the 1996 election in which the baseball bats had supposedly been out for Paul Keating since the ‘recession we had to have’ (43%).

The Australian Election Study comes up with generally lower, but still significant, proportions of late deciders. They ask ‘when did you decide how you would definitely vote in this election?’, and adding up the answers from ‘about the time the election was announced’ to ‘election day’ we get 38% in 2004, 41% in 2001, 51% in 1998, and 38% in 1996.

But does this exaggerate the impact of the campaign? There are a couple of reasons for thinking that we should put a lot of emphasis on the ‘finally decide’ in Newspoll and the ‘definitely vote’ in the AES.

Before the election, Newspoll also asks voters how certain they are to vote the way that they have just indicated. It finds that only a fairly small proportion of voters are still very undecided. In the final poll before every election since 1996 a remarkably consistent 9% have agreed with the proposition that ‘you could vote this way but there is just as much chance that you will vote for someone else’. A larger minority, of (depending on the election) 22%-31%, agree with the proposition that ‘you will probably vote this way but there is a slight chance that you may vote for someone else’.

In the last couple of elections Newspoll has also asked the certainty question in the months leading up to an election. The highest proportion of voters saying that they could go either way was 18% in June 2001, five months before the election. It was down to 13% early in the 2001 campaign and 10% early in the 2004 campaign. These numbers suggest that the overwhelming majority of people have made a certain or provisional decision before the campaign starts.

There is other political science research consistent with these Newspoll findings. In the Australian Election Study, more than 80% of voters nominate a party when asked: ‘Generally speaking, do you usually think of yourself as Liberal, Labor, National or what?’ The AES also asks people who they voted for at the most recent election and who they voted for the election before that. In the last two surveys, just over three-quarters of respondents who voted at the previous election say they voted for the same party both times. (Though new voters are likely to be less stable in their political preferences.)

The proportion of voters who rely on the campaign for their decision, then, isn’t 50% or anything close to it. It’s probably not even 20%, once we take out those whose vote is a random marking of the ballot paper intended to avoid being fined for not voting rather than to give a verdict on the campaign. Another 30% are probably going to vote according to a choice already provisionally made, often influenced by their general support of a party, but have not made a final or definite decision.

Of course the campaign is still very important – not just to persuade the very uncertain, but for the parties to ensure that supporters who are not rusted on do in fact vote for them on election day. But the electorate is not as volatile as people might think, if they just looked at surveys on when the final voting decision is made.

11 thoughts on “When do voters make their decision?

  1. Andrew, you are right. People who say that 50% make their choice in the last month are just silly. One only needs to look at how consistent are voting patterns in all but a minority of electorates (called marginal) so see elector’s views and allegiances are fairly stable.


  2. Andrew – I suspect you are right in your analysis, The Newspoll question turns on the word “finally”, as in finally decided. I suspect that many people have a tendency, perhaps a strong tendency to vote for a party, but at the same time reservations about some elements of what that party stands for, or the personalities within the party. Hence in prospect they indicate a strong likelihood to vote a particular way, and in retrospect a late final decision.

    Andrew Leigh has already pinged me for that paragraph in my post. I should explain, I was not seeking to make statement about the importance of the campaign. I was seeking to explain some of the contributors to volatility in the polls.

    In retrospect, I may have over-reached in using the broad Newspoll statistics to support my argument. Perhaps the more meaningful figure for my argument would have been the AES 20% who do not align themselves (even provisionally) to a particular party. It is these people (and perhaps others with a weak party alignment) whose indeterminate (perhaps flip-flopping) state between elections contributes to the volatility of the polls.

    Let me finish with an anecdote. Last election, someone in my workplace told me on the Friday night immediately before the election that s/he planned to vote Labor. On the Monday morning immediately after the election, s/he told me that s/he had voted Liberal. This sort of person, who can change their mind within 24 hours and largely on whim, is a nightmare for pollsters.


  3. Bryan – Your substantive point was correct – there are flip-floppers who make life very difficult for pollsters. I was just putting an argument as to why there are fewer of them than many people believe. I saw Andrew L’s comment, but I think he understates the between-election ‘volatility’ while again being substantively correct that most people vote consistently – which is why we get the stability that Boris noted. Another point to keep in mind is that volatility in the actions of individual voters does not translate into election day results volatility on the same scale, as to some extent voters changing in opposite directions cancel each other out. I wrote a post on this last year.


  4. Actually, I’m not sure that variability amongst voters and the stability of an electorate have much to do with each other. One is to do with the means and the other the SD (or whatever measure of spread you like). It would be quite possible to have quite decent spreads in many blue ribbon electorates, but as long as there is a big enough group of people that vote one way, it doesn’t matter. The amount that the means shift from election to election is probably a more worthwhile measure to look at (particularily when you get landslides for one party, in which case most of the “undecided” voters are voting one way, so you should be able to get an estimate of the real number that are willing to change parties, vs. the number that claim they do in polls).


  5. Conrad, I disagree with your first sentence but agree with the last one. This is what I meant: if you see in successive elctions that the overall voting pattern changed by only, say 0.1%, this means that few voters changed their mind. Of course, it could be that 25% changed their vote, but this was precisely cancelled by another 25% whose vote changed in an opposite way. But statistically this is extremely unlikely.


  6. I wondered a bit more about that Boris, so I looked at the relationships between the absolute swing with the absolute liberal/labor vote (i.e., whether electorates evenly split had a bigger swing) in the recent Vic elections. I chopped out the greens. THe overall swing wasn’t huge, so there could be better data sets to look at.


    We might expect a negative correlation if more marginal electroates have greater swings (i.e., the closer to a 50-50 split, the bigger the swing will be). There is such a relationship , but its basically just noise r = -.11.
    The distribution looks a bit better than the number suggests, with all the extreme values in the right places, which suggests to me that there are few very blue ribbon electorates that don’t have many of these voters, but as for the rest, the swinging voters are nicely distributed.


  7. It has taken me two years to master the crostabulation of the AES studies. They are the only ones worth looking at for study on national voting behaviour.
    Also, as the campaign for 2007 started about a month ago, the traditional questions about voting behaviour due to ‘the’ campaign are becoming more irrelevant. It is great to see this being discussed as I study the value of preferences and they hype behind so called ‘backroom deals’. The campaign will be running hot this week in APH, now that’s entertainment!
    with respect


  8. “Also, as the campaign for 2007 started about a month ago, the traditional questions about voting behaviour due to ‘the’ campaign are becoming more irrelevant.”

    Lisa – I don’t agree with this. Certainly, as I argued in the post, most people have already largely made up their minds before the campaign. But the campaign is important for reaching those with little interest in politics who will not being paying too much attention to who Kevin Rudd has lunch with or what shares Santo Santoro owns, even if the political class is whipping itself into a small frenzy.

    If you do an AES cross-tabulation, you will see that 40% of those who claim their interest in politics is ‘not much’ or ‘none’ say they decided their vote in the last week, compared to 13% of those who say have a ‘good deal’ of interest in politics. 70% of those with little interest in politics say they paid not much or no attention to election news on TV.

    These people are the know-nothings, which is why the campaign is important because it uses other ways of communicating with them: advertising in TV shows they do watch, direct mail, telephone. These methods are only occasionally used outside the campaign.


  9. Not too sure about that last post, Andrew. You could not possibly vote for Labour this election. (NSW) And the Libs? More or less invisible.
    What is left? I haven’t seen any publicity in our electorate from whatever/whoever else is running.
    Therefore I will decide when I see the names on the form.
    First vote will go to whoever I think least likely to recoup their deposit. Libs/Labour, Labour/Libs will go on the bottom.
    When the choice is disaster vs disaster there is no point in being interested.
    However a question: the polls are predicting a Labour win. By any rational measure Labour is unelectable. And while the Libs are not offering anything, they do not come across as that much worse — how could they be?
    Is what we are seeing the beginning of the reaction against anything branded Liberal?


  10. Peter hindrup wrote:
    Is what we are seeing the beginning of the reaction against anything branded Liberal?

    Not by a long shot. The NSW liberals need to find a moderate leader like Brogden – he’d have romped it in. They have made themselves unelectable with branch stacking tactics reminiscent of the worst of Labor and are now paying the price.


  11. “These people are the know-nothings, which is why the campaign is important because it uses other ways of communicating with them: advertising in TV shows they do watch, direct mail, telephone. These methods are only occasionally used outside the campaign.”

    TV political advertising in the weeks leading up to the election is probably targeted towards the “know-nothings” – which explains its often blunt crassness.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s