Essential Report Microblogging Disclaimers

Frequently I post federal polling results and calculations to the microblogging sites Twitter (X) and Bluesky (links to my accounts in sidebar); here is an example I posted today:

The standard items I include are the primary votes for all parties, any two-party preferred or Labor vs One Nation "shadow 2PP" issued by the pollster, and what I call "my conversion".  "My conversion" is a 2PP figure I get by applying my own estimate of 2025 election preference flows to the primary votes released by the pollster, and is the figure I use in my Labor vs Coalition 2PP aggregate and my Labor vs One Nation 2PP estimate, both of which can be seen on the sidebar together with a link to the methods page for them.   I have been including primary vote changes recently because of interest in whether One Nation is going up or down, but probably won't keep doing so for long.  (In fact the one shown has a typo in it, the Greens are down 1 not up 1).  

In the case of one particular pollster, Essential Report, I have to add a lot of disclaimers because of the unusual properties of that poll.  And I'm tired of typing them out so often so I have decided to put them in article form to link to in the future to save the time of doing so.  If this has the effect of causing some changes in some of the less credible aspects of Essential's polling, I will take that too.

So here we go:

1. Different Scale: Every other significant pollster publishes primary votes with undecided respondents in some way removed or reallocated.  However, Essential does the following: if a voter picks "Don't know" on the first voting intention question they are asked which party they are leaning towards.  If they again pick "Don't know" they are left in, creating an undecided figure that is usually 5% or 6%.  

In theory this should mean that the parties all get slightly lower primary votes from Essential than on average from other pollsters.  In practice this doesn't happen because of:

2. Very Low Independent/Other readings:  Essential measures the independent/Other vote via a voting option of "Independent or other party" which appears near the bottom of the voting options list.  For whatever reason this attracts way lower numbers than other pollsters' independent/other options - numbers too low to be believable.  The average for the first half of 2026 has been 5.7% (6.0% after reallocating undecided) compared to 10.9% from all other pollsters.  This category polled 15.1% at the 2025 federal election, although it is very likely that other pollsters are right in finding something like four points of that has since moved to One Nation.

3. Unusual 2PP Method:  Essential calls its Labor vs Coalition method "2PP+".  Instead of presenting a two-party figure that sums to 100 such as 51-49, Essential leaves undecided voters as undecided in the 2PP so the figure might instead be 48-46 with 6 undecided.  Essential started doing this after the 2019 election polling failure, arguing that in a close election it was necessary to stress that a slim lead was not conclusive and undecided voters could yet cause the trailing party to win.  As I pointed out at the time this reasoning placed too much weight on the apparently spurious excuse that undecided voters caused polls to be wrong in 2019.  The 2019 failure was more likely caused by bad sampling, overly simplistic weighting practices and herding or a herding-like phenomenon.

4. Aberrant Respondent Flows: Essential's 2PP+ is largely based on asking voters who support minor parties which of Labor and Coalition they would preference highest; if the respondent then says they don't know then the last-election preference split for that party is used.  This is basically a respondent preferences method but for whatever reason it produces flows to the Coalition that are stronger than the respondent flows from other pollsters (how much stronger varies) and much stronger than last-election preferences.  On average the Coalition in the 2025-8 term is doing around two points better on Essential's "2PP+" (with undecided removed) than it is on my last-election estimates.  At the same time, Morgan's respondent preferences on average have the Coalition doing one point worse than last-election preferences - something that there might actually be a reason for if One Nation are taking voting intention from the right-wing end of the Others pool.  These preference flows in Essential are so odd that they should have a careful look at their operations (in particular spreadsheet calculations and whether their panel adequately controls speeding respondents) to see if they are doing anything wrong.

5. House Effect: Even after using last-election flows for Essential to negate the impact of its weird respondent flows, Essential still has better numbers for the Coalition overall than other regular pollsters - it tends to get the Coalition's primary on the high side and the Greens on the low side compared with the average.  This doesn't mean it's necessarily wrong but is worth keeping in mind when I post conversion figures for it that are lower for Labor than those for other regular pollsters.  For the term so far, Essential is running 1.3 points better for the Coalition on my 2PP conversions than my aggregate at the time of its releases.  Some less regular pollsters (Freshwater, Spectre and Fox&Hedgehog) may have similar properties but have only released a few polls each.  

I also note that Essential doesn't have a good accuracy record in recent years.  It was at the bottom of my accuracy tables for federal 2019, federal 2022 and the Voice 2023 and 8th out of 10 (ahead only of Freshwater and the once-off Ipsos) for federal 2025, which is not to say it cannot perform well in the future or should be entirely ignored.