Blogging Senate forecasts and results in the WA Senate re-election until officially declared.

Twitter: @AU_Truth_Seeker


Tuesday, 1 April 2014

(UPDATED) Private data

I thought of writing a post predicting the election of both Group C independents as an April Fool's day joke, but then I actually got enough information to write a real post based on new information.

Exclusive private polling emailed by a white knight to TruthSeeker HQ has revealed that:
- The Palmernami is seemingly true, with a high and rising vote for the billionaire's party. This further backs a vote that may reach 10% or higher!
- Labor Vote is weak, languishing below it's sickly low 2013 vote
- Green vote is strong - at levels at or slightly above one quota
- Liberal vote is inevitably down
- Not much decrease for the critical "others" block.

Primary votes (UPDATED)
There are quite a number of organisations who have the interest and funds to undertake polling for this election. Most will run multiple polls over a period of time, and the necessity of private polling would also be driven by the lack of a reasonable free alternative, unlike for general elections.

Methods will vary and the pollsters conducting these polls will vary too - the best approach is to use as much data as possible and allow for a trend. This best practice approach is routinely used by Will Bowe (Poll Bludger), Kevin Bonham and others.

We will never know if a party with a vested interest is selectively releasing information to push a particular case. Alternatively, such data may be the result of a push poll or be just wrong.

In this case, I have confidence in the data provided to me as I have also seen the text of all questions and I can confirm the questions were unbiased and fair.

Pulling all this together, I ran two different scenarios based on data provided to me - a moving average and a point estimate.
The point estimate I used was:
LIB/NAT: 40%
ALP: 26%
GRN: 14%
PUP: 8.5%
OTH:  11.5%

Of course, these come with the standard disclaimer about margin of error, etc.

Beyond this, I can't say whether it was a "right" or "left" party - that would be too obvious.

Monte Carlo Simulations
Before running this through my Monte Carlo engine, I modified the percentage variation to a lower 5% level - this gives more weighting to the polling with which I've been given access to.

Very likely: (100% likely)
LIB: 2
ALP: 2
GRN:1

Possible:
LIB3: 49%
PUP: 51%

Wider picture
This is the first polling I've seen that directly relates to the senate election, so it is appealing that such polling is occurring. But we need to consider each poll's performance in relation to other polls and other information (whether public or private). We also will learn the accuracy of senate polling as this election progresses - to date we're untested as to whether specific senate polling is accurate for a specific senate election.

This modelling backs up previous modelling that the election of PUP is likely to offset the elction of the 3rd LIB senator.

5 comments:

  1. This is a bit of a tease, no doubt you are sworn to secrecy about the source but could you at least state the assumed primaries...?

    ReplyDelete
    Replies
    1. I know - I feel a bit bad. But I am happy to post some more accurate data just I'll have to do it tonight.

      Delete
  2. Was it a left or a right source? How do you think a potential low turnout will affect/benefit certain parties?

    ReplyDelete
  3. I was robopolled tonight, asked to choose between Lib, ALP, Greens, PUP, OTH. Asked about whether I'd seen Greens advertising on TV or radio, leadership approval, demographic information and possibly something else. I can't remember who they said they were, but it wasn't one of the majors.

    ReplyDelete