The latest ipredict snapshot is out, showing supposed party support as at an unknown recent time (update – it is a random time but released later, this week it was at 9.30 am on Thursday). There’s a major problem with this approach.

I’ve been watching party support on ipredict over the past week and it has been fluctuating wildly at times.

Last Friday there was a major push upwards in Green support, and a signifcant push down on the other small parties. Things settled back to new levels obver the weekend.

On Monday Act had a major push up.

Today there was another Green rise and the other small parties were hammered down. During this volatility the snapshot was taken. Forecast party vote shares are now:

National 50.0% (49.0%)
Labour 28.1% (30.2% )
Greens 10.3% (8.0%)
Act 2.0% (4.1%)
New Zealand First 3.6% (4.0%)
Maori Party 1.4% (1.5%)
Mana Party 1.4% (1.0%)
UnitedFuture 1.0% (1.6%)
(last week in brackets)

Those positions and changes are what is reported but it pays to look at the approximate ranges (which can be extreme in one day, particulalry notable last Friday, Monday and today).

National: 47.5 – 54.0
Labour: 25 – 31
Greens: 7.8 – 11.6
Act: 2.0 – 5.7
NZ First: 2.6 – 6.4
Maori Party: 0.1 – 1.5
Mana Party: 1.0-1.6
UnitedFuture: 0.4 – 1.8

It’s obvious that opinion is very unlikely to swing that much, in some cases several times in a week.

It’s far more likely that parties are playing the pools to paint a poll picture. It looks as unreal as modern art. Check the graphs to see for yourself.

UPDATE: swings around the time of the snapshot:

5.7% Wednesday evening
4.3% Thursday 8.58 am
2.1% 9.09 am
2.0% 9.30 am snapshot
3.1% at 1.32 pm
3.5% at 5.50 pm

Green Party
9.0% at 8.55 am Thursday
9.8% at 9.07 am
10.3% at 9.30 am snapshot
10.0% 12.30pm

3.3% at 8.55 am
3.6% at 9.30  snapshot
4.3% at 10.30 am

United Future
1.7% at 7.13 am
0.4% at 9.21 am
1.0% at 9.30 snapshot

Leave a comment


  1. Pete, how did you rule out events/new information as a source of price movement? Volatility is a feature not a drawback if this explains it.

    It is easy to see and ignore momentary spikes produced by one-off large trades on the site. They are quickly corrected by the crowd. Your ranges are smaller without the large, obvious, momentary spikes.

    Volatile compared to what? You’re certain at least as much volatility wouldn’t emerge in polls with margins of error at 3% if polls were run daily or hourly?

    If you doubt comparative accuracy we can observe it after the fact: at the last election iPredict was more accurate (closer to the final result) than 15 of 19 polls published between our launch and election day. Similar success ratio for other markets overseas for 20 years. No reason to expect a different performance this time.

    Some evidence on overall performance here: https://www.ipredict.co.nz/app.php?do=ipredict_perf

    Re: “an unknown recent time” In the update we say: “This week’s [snapshot] was taken at 9:10am today.” We deliberately randomise the time to avoid any risk of gaming.

    • Matt, I don’t doubt that it becomes more accurate closer to the election, but taking a single snapshot point now while there is such significant fluctuation makes the reported positions a lottery. Perhaps there’s a reason why other polls take their numbers over several days.

  1. | Your Dunedin

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s