As the party settles in under a new leader who remains largely unfamiliar to the public at large, the Greens have more reason than usual to keep a careful eye on their standing in the polls.

Given recent breakthroughs at state elections in New South Wales and Victoria, and the mounting unpopularity of the two major party leaders, the party could certainly be forgiven for feeling bullish about its long-term prospects.

However, the latest batch of results to emerge from polling over the weekend, while telling a consistent story on two-party preferred, paint an extremely confusing picture with respect to the Greens.

The best news for the party comes courtesy of Ipsos, which has the Greens’ share of the vote soaring to 16% — two points higher than the previous poll, which was already its best result since Ipsos began its monthly series for the Fairfax papers in November.

But it was a very different story from yesterday’s Essential Research poll, which had the Greens at only 10%, down one point on an already modest reading the previous week.

The gap goes beyond anything that might be accounted for by normal polling variability.

The commonly cited margin of error for opinion polls with standard sample sizes of a bit over 1000 is 3%, but this presumes a roughly even division of support between two available options.

In gauging support for a minor party that polls in the low double figures, the theoretical error margin is actually more like 2%, if not lower. That leaves a big chunk of daylight between the lower end of Ipsos’ range and the upper end of Essential’s.

The principle behind the margin of error is that 95% of a pollster’s results should fall within its range, and it’s certainly possible that one of the two pollsters — or, less likely, both — have turned in that one poll in 20 that strays outside.
Ipsos makes this relatively easy to diagnose by providing detailed breakdowns of its results by location, gender and age, which allows for identification of errant results within individual sub-samples.

On this occasion, the standout result is that the Greens outperformed the Coalition by 31% to 28% among the 18-24 cohort, and came in only slightly below the 33% support recorded for Labor.

While this is no doubt excessive, the occasional quirky result is only to be expected when a poll is broken down into very small components of fewer than 200 respondents — and in this case, the 18-24 age group accounts for barely 10% of the overall voting population. The anomaly, if that is indeed what it is, is unlikely to explain more than 1% of the overall result.

Another possible explanation lies in the pollsters’ different modes of surveying, with Ipsos contacting respondents at random by telephone, and Essential Research drawing samples from an online panel of over 100,000 volunteers.

It might be tempting to propose that some sort of inversion of the “shy Tory” effect was at work, with “loud Greens” being particularly eager to advertise their political correctness to live interviewers, but less motivated to do so in the anonymous context of an online survey. But this hasn’t been borne out by Essential Research’s polling in the past, which landed right on target with respect to the Greens in its final poll before the 2010 election, and came in 1.4% too high in 2013.

Instead, it appears that Essential Research has been steadily peeling away from the remainder of the pack since the election of the Abbott government. As illustrated in the chart below, this has reached new heights since its rivals began to detect mounting support for the Greens as of late May, while Essential remained locked within its long-established track at around 10%.

This pattern had a parallel in polling conducted in Britain between November and February, when phone pollsters recorded a spike in support for the Greens that seemed to escape newly established online polls — though not the industry-leading online pollster, YouGov.

By the time the election came around in May, whatever effect the phone pollsters were tapping into had dissipated, and all modes of pre-election poll sang from the same song sheet in having the Greens at between 4% and 6% (slightly excessively, as it turned out).

As with so many things in life, it seems likely that the truth of the Greens’ current level of support lies in the middle of the extremes as recorded by Essential and Ipsos.

Indeed, evidence to this effect was furnished by the third poll conducted over the weekend — namely, the very first Newspoll survey for The Australian under the brand name’s new management, Galaxy Research.

As conducted by Galaxy, Newspoll now includes online as well as phone polling, which may help explain why its result for the Greens split the difference between Ipsos and Essential by coming in at 13%.

This is also well in line with the latest finding of the BludgerTrack poll aggregate featured on my blog, The Poll Bludger, which puts support for the Greens at 13.8%.

By any standard other than that set by the latest Ipsos result, this is highly encouraging for the party, being a 5.2% improvement on its disappointing showing at the 2013 election, and 2% clear of its previous record in 2010.