Two polls, two governments
.
.
Two polls out recently give completely different outcomes if an election had been held over the last week or so.
One, the Roy Morgan poll would result in a change of government – whilst the Herald Digi Poll would (without overhangs) allow National to almost govern on it’s own. The results,
.
|
Roy Morgan Poll |
Herald Digi-poll |
|
| National |
43.5% (-4%) |
48.5% (+1%) |
| Labour |
32.5% (+2%) |
36.4% (+4.4%) |
| Greens |
13.5% (+1%) |
9% (-1.7%) |
| NZ First |
5% (+2%) |
2.5% (-3%) |
| ACT |
0.5% (n/c) |
0.1% (-0.1%) |
| Mana |
0.0% (-0.5%) |
0.5% (+0.2%) |
| Maori Party |
2% (-0.5%) |
1.1% (-0.4%) |
| United Future |
0.5% (n/c) |
0.0% (-0.3%) |
| Conservative Party |
2% (n/c) |
1.3% (-0.1%) |
| Undecideds/Wouldn’t Say |
5% |
11% |
(n/c = No Change)
.
Two polls, two outcomes, two governments. So which is more accurate?
In a previous blogpost (see: Three recent polls), a comparison was made between Roy Morgan, Colmar Brunton, and Ipson Poll. Of the three, Roy Morgan was closest to actual election day results in 2011.
So let’s compare Roy Morgan; the DigiPoll, and Election Day results,
.
|
Roy Morgan 24 Nov 2011 |
Digi Poll 25 Nov 2011 |
2011 Election results |
Closest Polling result |
|
| Right bloc: | ||||
|
National |
49.5% |
50.9% |
47.31% |
Roy Morgan |
|
Maori Party |
1% |
0.4% |
1.43% |
Roy Morgan |
|
ACT NZ |
1.5% |
1.08% |
1.07% |
Digi Poll |
|
United Future |
0.5% |
0.0% |
0.6% |
Roy Morgan |
| Left bloc: | ||||
|
Labour |
23.5% |
28% |
27.48% |
Digi Poll |
|
Greens |
14.5% |
11.8% |
11.06% |
Digi Poll |
|
Mana Party |
0.5% |
0.3% |
1.08% |
Roy Morgan |
| Other: | ||||
|
NZ First |
6.5% |
5.2% |
6.59% |
Roy Morgan |
|
Conservative Party |
n/r |
1.3% |
2.65% |
Digi Poll |
.
Roy Morgan was slightly more accurate than the Heral Digi Poll.
Interestingly, Roy Morgan seems to be the most accurate pollster when it comes to National, beating Herald DigiPoll, Colmar Brunton, and Ipsos.
Equally important to Roy Morgan’s polling for preferred Party, is polling for Satisfaction/Dissatisfaction for the government of the day – in this case, National.
Roy Morgan asks respondants,
“Generally speaking, do you feel that things in New Zealand are heading in the right direction or would you say things are seriously heading in the wrong direction?”
The results seem to back up National’s fall in preferred Party stats,
.
| New Zealand Roy Morgan Government Confidence Rating (Government of John Key): Interviewing Dates |
|
Jan 30-Feb 12, 2012 (post election) |
Jan 2-13, 2013
|
Jan 14-27, 2013
|
Jan 28-Feb 10, 2013
|
Feb 11-24, 2013
|
Feb 25-Mar 10,
|
|
| Right direction |
57% |
53.5% |
57% |
55% |
54% |
51.5% |
| Wrong direction |
30 % |
33.5% |
30.5% |
30.5% |
32.5% |
37.5% |
|
Roy Morgan GCR# |
127 |
120 |
126.5 |
124.5 |
121.5 |
114 |
| Can’t say |
13% |
13% |
12.5% |
14.5% |
13.5% |
11% |
| TOTAL |
100% |
100% |
100% |
100% |
100% |
100% |
#Roy Morgan GCR = Roy Morgan Government Confidence Rating (The Roy Morgan GCR is 100 plus the difference between the percentage of New Zealanders who say the country is “heading in the right direction” and the percentage who say the country is “seriously heading in the wrong direction”).
.
Acknowledgement: Roy Morgan Poll
Soon after the 2011 election, National rated highly with respondants, with 57% approval. Since then, except for a ‘blip’ at the beginning of the year, National’s approval rating has dropped from 57% to 51.5%.
Conversely, those expressing a view that National was headed in the wrong direction, rose from 30% soon after the 2011 election to 37.5%.
Those who Couldn’t/Wouldn’t say have dropped from 13% to 11% – meaning that people’s views on National are firming up – and becoming more pissed off.
Once Mighty River Power is part-privatised, expect to see National’s support plummet even further.
As this blogger has been predicting consistently; we will see a change of government in 2014 (if not earlier).
.
*
.
Previous related blogpost
Census, Surveys, and Cellphones
References
Final poll: Nats win looks certain, Winston over 5% (25 Nov 2011)
Roy Morgan Poll (18 March 2013)
Labour rises at expense of allies (21 March 2013)
.
.
= fs =
l’d like 3rd option of neither, although roy is close
Hi Frank. Your analysis to determine poll accuracy assumes each ‘party result’ is statistically independent. However they are non-independent (ie, an increase/decrease for one party results in changes in the other).
From a statistical point of view, I think it makes more sense to determine accuracy by calculating the ‘total error’ in a poll’s results. What could be REALLY interesting with this approach is that you could calculate ‘total error’ per block of parties – and this would tell you whether a poll is more right/left leaning etc.
Not trying to by annoying – I just think there’s lots more interesting stuff that you can get from that sort of analysis.
Thanks for that, Andrew. You’ve made valid points, and your blogposts on this issue (on your blog) also add to the overall picture.
Re your comment,
I’d also assume that Party results depend on the over number of respondants and the number of Don’t Knows/Refuse To Answer group.
It’s why I connected Roy Morgans Party Polling with their Government Confidence Rating – they both seem to indicate a fall in support for National. So if one had gone up whilst the other showed a drop – that’d give us a confusing picture overall.
Food for thought on that point.
Thanks Frank – Nice to know my blog is interesting to someone
Yeah I’d agree with you that Government confidence ratings would be correlated with party support. Actually they are really interested because they could indicate ‘softening’ or ‘hardening’ of support for a party.
The thing with ‘who would you vote for’ questions is that they are ‘all or nothing’. We might find, for example, that 38% of people support a party, but this doesn’t mean that all those people support it particularly strongly. Some people may support a particular party simply because they see no viable alternative.
The same goes with some policy-type questions. By asking a yes/no question we may find that 60% are against a particular policy, but this doesn’t give any indication of how strongly they are against it. For that, you’d need a different type of question.
The one that they do on the Yahoo! New Zealand News website yesterday did not even have New Zealand First listed as a party, even though I am fairly certain it is the fourth largest party in Parliament and that over 100,000 people voted for it in 2011.
Yeah – that plus self-selecting political polls on website are generally not that useful.