Kiwiblog

A reader writes:

I have been watching every stuff poll that has something to do with a political topic over the last few weeks and every one has started out one way and then low and behold 12 hours later it has completely skewed the other way.

This is not surprising. I have blogged on this stuff several times previously. The media need to be more pro-active in not just detecting and preventing fraud, but also in exposing those who do it. Even though these polls are unscientific, they are reported on and seen by many people.

Generally the polls all start out favouring the nats and then magically skew the other way, and not by a little, were talking massive swings. Now from a statistics point of view, this is not generally possible.

Yep – this is a pattern, always the same way.

Here is an example that i have emailed into stuff.co.nz today. I live in London, and last night the poll “National MP Lockwood Smith’s comments about Asians and Pacific Islanders were …” was dramatically on the side of ‘Fair Enough’ and ‘Undiplomatic but no big deal’. So i get to work this morning and have another look at the poll and low and behold it’s completely flipped again.

So i thought i would take some sample data:

Hard data is always useful.

At 12:20 am (NZ Time)
Racist (12485 votes, 56.7%)
Fair enough (3893 votes, 17.7%)
Undiplomatic but no big deal (5653 votes, 25.7%)

At 2:20am (Nz Time)
Racist (15215 votes, 61.2%)
Fair enough (3922 votes, 15.8%)
Undiplomatic but no big deal (5715 votes, 23.0%)

Difference
Racist +2730
Fair enough +29
Undiplomatic but no big deal +62

And here we are at 2:45am, when most kiwis are snug in bed and the ‘Racist’ side is still climbing a a rate of a vote every second or so.

That is absolutely damning. Fairfax should not just take measures to eliminate those fake votes, they should expose the IP address the multiple votes are coming from.

Now i know that stuff polls are not scientific in any way but if someone is doing this intentionally and i believe they are they are trying to skew public opinion because if they see that 70% of the population is voting a different way to them they will start to think that maybe they are wrong. There are a lot of people that look at these polls and in an election where every vote counts, if only a few change their votes because of ’supposed’ public opinion, then we have a big problem.

The NZ Herald had to fix up their polls a little while ago because of tampering, so now it looks like stuff should do the same.

The solution is not that difficult.  Fairfax staffers should be monitoring their polls, as this reader has done, and take action when they detect obvious fraud.

I sometimes have people try to do bulk multiple voting on my own blog polls. Generally I notice it within an hour or two, and reverse it. And I expose where it was done from – for example a VUWSA computer was used one time to do it.

A quick follow up. It’s now 4:20am NZ time and i have taken the next sample. This data is from 2:20am – 4:20am

At 4:20am
Racist (18051 votes, 65.1%)
Fair enough (3934 votes, 14.2%)
Undiplomatic but no big deal (5742 votes, 20.7%)

Difference from 2:20am
Racist +2836
Fair enough +12
Undiplomatic but no big deal +27

Beyond any doubt, And I just had a look myself at 6.40 am and the growth in the last two hours has been:

Racist +3080
Fair enough +6
Undiplomatic but no big deal +20

The reader who sent this to me has also e-mailed Fairfax/Stuff. I’ll publish any response from them.

My latest blog poll, which like all of them is fun not a scientific poll, is asking who people would vote for. As one would expect with my political leanings, there was a lot of National and ACT voters – around 75%.

But then as I was out enjoying a night on the town, I got a text mentioning that someone appeared to be stacking the poll and multiple voting for Labour and the Greens. I was not surprised as a near identical thing had happened almost a dozen times previously with other online political polls, as I last blogged about here.

I had set it to stop multiple voting on the basis of cookie, not IP address, so that more than one person in a large workplace (such as Parliament) could vote. But I did have it set to log voting by IP address, in case anyone set up a bot to automatically delete the cookie and multiple vote. Or if someone is a total loser, in case someone sat by a computer for a few hours manually deleting their cookie, and multiple voting.

So what do the logs say. Let us look at votes for Labour:

Out little poll hacker started work at 5.20 pm and was voting several times per minute. Votes for the Greens were also being cast during that time, so it was maybe six to eight votes per minute – almost certainly automated.

And they look to have stopped around 7.19 p.m.

In total they cast around 600 votes for Labour and 650 votes for the Greens.

Now VUWSA is the Victoria University of Wellington Students’ Association. They take over one million dollars a year in compulsory fees from Victoria students. Students will be no doubt pleased to hear the sort of activities they are forced to fund now includes stacking and multiple voting in online polls.

I hope Salient will make some inquiries of the Association, specifically which (presumably) executive or staff member was using the computer with that IP address. How many times have they done this before? Do they think it is compatible behaviour with being custodians of compulsory fees?

And their colleagues around the country will no doubt thank the person responsible for adding another example to the armory of those who support voluntary membership of student associations.

UPDATE: The VUWSA IP address seems to include the Salient computers, so maybe Salient should also interview its own staff or volunteers on who was using it.