The New Zealand Electoral Commission has now released its data showing votes in the 2011 electoral-system referendum at the level of the polling place.¹

I have taken a very quick look at the data, and can report some interesting correlations. Keep in mind this is aggregate data, albeit at the level of the polling place, and not individual-level data.² All of the numbers reported here are thus simple correlations of the vote percentages for a given option on Part A of the ballot (keep or change from MMP) with percentages on Part B (the choice of an alternative system, should “change” win).

Keep, correlated with:
FPTP (FPP), -.543
AV (PV), .251
STV, -.019
MMM (SM), -.574
Informal, .745

(Acronyms in parentheses indicate how the systems were identified in the referendum, which is not always consistent with how they are more commonly known outside NZ, and to the terms I prefer to use.)

These correlations suggest that voters who were happy with the current MMP system were more likely to simply abstain than to select any replacement option from Part B. This is a trend that, as noted earlier, was already apparent from the national aggregate data.³

It is interesting that the only electoral system from Part B that has a positive correlation with a polling place’s percentage vote to keep MMP is the Alternative Vote (known in New Zealand as “Preferential Vote”). It is somewhat surprising that pro-MMP voters would not be more likely to have chosen the only other proportional system on offer, STV. Yet the vote for STV bears hardly any correlation at all with “keep MMP”.

Now, correlations between “change” (i.e. replace MMP) with Part B options:

Change, correlated with:
FPTP, .531
AV (PV), -.254
STV, .034
MMM (SM), .594
Informal, -.751

Not surprisingly, these are close to mirror images of the correlations with “keep MMP”, except for the STV vote (which is simply not well correlated with however a voter voted in Part A).

It is evident that voters who were discontented with MMP were unlikely to abstain from choosing an alternative, and were about as likely to go for either of the majoritarian systems on the ballot that did not entail ranked-choice ballots–that is, FPTP or MMM (SM), but not AV (PV).

As for the small percentage of voters who declined to mark a choice on the keep/change question, this is positively correlated with only one option in Part B: FPTP. It is a weak correlation, at .145, but all of the others are negative and not much different from zero. Perhaps this means that the sort of voter who would abstain in Part A was a low-information “good old days” FPTP voter. (“I don’t get this whole electoral systems thing, but I know things were better before we got rid of FPP.”)

There are more detailed analyses one could do, particularly taking into account partisan trends in electorates and polling places,⁴ but this makes for a good starting point.

_______________
Small corrections, 2 Feb.

1. It is in section 7 at the Commission’s Statistics page. I am grateful to Robert Marsh, Senior Project Leader at the Electoral Commission, for sending me the link. [↩]
2. There are 6,066 observations. The median number of referendum ballots cast per polling place was 129. [↩]
3. However, as Manuel Alvarez-Rivera noted in a comment to our earlier discussion, at the electorate (district) level, the correlation between “Keep MMP” and an informal vote (abstention or invalid) in Part B, was 0.97. So, when we look at the polling place, the correlation drops quite a bit. [↩]
4. The latter would be very time-consuming, as the Electoral Commission does not seem to have put out a single data file that contains all the polling places, but only the electorate-by-electorate data. So it would be very time consuming–albeit possible–to do this analysis. [↩]