Making sense of the polls

Making sense of the polls

Some More Poll Details

Follow the PM’s Approval Rating, not the Preferred PM

If the election is competitive, the Preferred PM measure will invariably show the incumbent with a big lead over the opposition leader. The opposition leader only gets a preferred PM lead when the Two Party Preferred (TPP) shows a landslide to the opposition. As a result, preferred PM isn’t a useful measure. A consequence of this is that it’s probably better to follow the party standings on issues, as opposed to the leaders‘ standings.

The PM’s Newspoll net approval rating (satisfied minus dissatisfied) is more important. Possum Comitatus has a post that relates the PM’s net approval to the government’s TPP. Possum finds that, while the PM’s net approval has always been important to the government, it has become critically important under the Rudd/Gillard government. That means that if Rudd has a net positive approval rating on election day, Labor should win. Unlike the PM’s approval, the opposition leader’s approval is effectively meaningless as far as TPP is concerned.

Rudd’s current Newspoll approval rating is 42% satisfied, 41% dissatisfied, for a net approval of +1. Usually about 10% will still say they have no opinion of a long-time incumbent, so around 7% still remains to be won here, and of course both satisfied and dissatisfied scores can change as new policies are announced, etc.

How is the Two Party Preferred Estimated?

Pollsters ask respondents their voting intentions, which are then listed as first preference votes for Labor, the Coalition parties, the Greens and all Others. From this, the pollster has to derive an estimate of the two party preferred vote between Labor and the Coalition. There are two methods for this.

  1. Use the preference flow data from the last election. At every election, the Australian Electoral Commission (AEC) calculates a national two party preferred (TPP) result between Labor and the Coalition. In the vast majority of seats the final 2 candidates will be a Labor candidate and a candidate from one of the Liberal or National parties. However, there are generally a few seats where an Independent or minor party candidate makes the final 2 candidate count. In these seats a special TPP count is conducted. Preference flow data from minor parties to major parties is released by the AEC, and at the 2010 election roughly 80% of Greens preferences and roughly 40% of all other preferences went to Labor. This data is used by the pollsters, adding 80% of Greens primary and 40% of all others to Labor’s primary vote to get Labor’s national TPP. This method is now the industry standard method.

  2. Ask minor party voters which major party they will give their preferences to. This method seems reasonable, but it has some drawbacks. For one thing, the total minor party vote in a poll will generally be around 20%. Asking these 20% for 2nd preferences introduces another form of sampling error into the poll. Another issue is that many minor party voters don’t have a clear idea of which major party they prefer. For these reasons, the preference flow data based on the last election is preferred.

Only one pollster now reports respondent-allocated preferences, and that pollster is Morgan. His multi-mode polls have had very big sample sizes of over 3500, and it is significant that Labor has done about 1% better since Rudd’s return using respondent-allocated than previous election preferences. In my poll tables I have used the previous election preference method, but Morgan’s last poll was 52-48 to Labor using respondent-allocated preferences, and only 50.5-49.5 to Labor using the previous election. If we were to add 1% to Labor to the current 50-50 polls, Labor would now be ahead, but we should be cautious in making this assumption given the drawbacks with respondent-allocated TPP estimation.

Polling in Australia is Easier than in Other Countries

Most countries which we are interested in have voluntary voting rather than compulsory voting; this applies to the US, UK and NZ for example. In these countries, pollsters must make assumptions about who is actually going to vote, then poll those who are thought likely to vote. Because Australia uses compulsory voting, pollsters can just use the demographic weights from the Australian Bureau of Statistics (ABS). In the case of landline pollsters, young people are hard to get, and so the pollster will keep calling until they have the required number of young people to match the state demographics.

A result of this is that, although there can be big swings in the final days of an election campaign, the final polls in Australia are pretty accurate. Antony Green reports on a Canadian state election, in which final polls had the opposition ahead by 9%, but the government won by 5%! This simply won’t happen in Australia. Although sometimes one poll is way out, the other polls will generally be about right. Newspoll hasn’t made a major error with primary votes at a state or federal election for a long time, so I would trust Newspoll to have its final poll right.