Oz Politics Blog

The editorial in today’s Oz was something else. It argued that an unusual single poll should be treated seriously. The core of the argument follows.

Mr Costello makes two mistakes. First is his rejecting out of hand Labor’s two-party-preferred slip from 52 to 50 per cent, on the basis that it falls within the poll’s margin for error, a comfort not afforded the Government in previous columns by Mr Costello justifying Treasury spokesman Wayne Swan’s disastrous strategy to reject government tax cuts in the 2005 budget. More importantly, Mr Costello errs in assuming the two-party-preferred measure is the one that matters. While he is correct that Australia has compulsory preferential voting for federal elections, the Newspoll preference allocation is artificial, based on the previous election result.

After many years of experience with Newspoll, The Australian knows the best indicative measures are primary voting intentions and preferred prime minister. Both rely on direct responses from survey participants. The party’s support fell to 37 per cent in the latest poll, while Mr Beazley’s personal rating has been low all year. In addition, last week’s poll was particularly significant because it showed support for Labor had slumped in the areas in which it should do best, health and education. It also showed that John Howard outrated Mr Beazley on economic management by almost three to one — 59 per cent to 20 per cent — despite rising interest rates. Contrary to Mr Swan’s view, Opposition leaders can poll well in economic management, as shown by Newspoll support for John Hewson and John Howard in that role.

The critical number for Mr Beazley is that Labor’s primary support has fallen below 40 per cent, to 37 per cent, considered an unwinnable position for Labor. A collapse in primary support to 40 per cent was the point at which The Australian’s political editor, Dennis Shanahan, accurately wrote off Labor’s chances under Mark Latham, declaring that its campaign had fallen into the “fatal zone”. At 37 per cent, Labor’s first-preference support is now level with that of the October 9, 2004 election, which rejected Mark Latham. The fall comes after the party’s first-preference support has bounced between 41 per cent and 42 per cent for each fortnightly Newspoll since August. Some of the blame can be sheeted home to the scandals surrounding state Labor during the survey period in NSW, Queensland, Tasmania and Western Australia. But this is of little comfort to Mr Beazley. As explained by Shanahan, it merely reflects the softness of support for Labor federally. At the first sign of trouble it collapses, and Labor’s negative campaigns on Iraq, IR and interest rates can be easily derailed. As explained by Graeme Morris on the ABC’s Lateline on Monday, Mr Beazley does not yet have the presence or the strength to overcome any sort of blip. Internal ALP concern at the poll result is heightened because, with Christmas and the election looming, nervous party members fear Labor is running out of time.

I happen to think Beazley is in some difficulty. While I think him electorally competitive, as I argued a bit over a week ago, my considered assessment is that Howard is more likely than Beazley to win the next election. My assessment is consistent with the betting market. It is also consistent with Morgan’s polling on who will win the next election. Morgan is the only polling company that asks this question, and fairly consistently since the 2004 poll the result has been 60% for Howard and 30% for Beazley.

I saw in today’s Age that I am not alone in my assessment that Beazley has some difficulty.

Nonetheless, today’s editorial in the Oz was psephological twaddle. In practice, opinion polls are much less reliable than mathematical theory suggests. Andrew Leigh and Justin Wolfers have noted, “the ‘margin of error’ reported by the pollsters substantially overstates the precision of poll-based forecasts. Further, the time series volatility of the polls (relative to the betting markets) suggests that poll movements are often noise rather than signal.”

Opinion polls must be considered in context: within the time series, in comparison with polls from other pollsters, and against the backdrop of political events. Often one must suspend judgment on a single poll that could mark a discontinuity, but in all likelihood is random noise. In my view, to read softness into this single poll is little more than isogesis.

While Labor’s primary vote has declined slowly over the last two months on all opinion polls, the most recent four point drop on Newspoll was anomalous. In my experience, most four-plus point jumps in opinion polls turn out to be random noise. They disappear as fast as they came. I expect Labor’s primary vote will improve in the next opinion poll.