New Zealand 2011: Party and electoral-system choice

(corrected and re-posted)

By using the polling place data posted by the New Zealand Electoral Commission ((And throwing aside all concerns about ecological fallacy!)) we can see the correlations between a party’s strength and voters’ electoral-system preferences.

The figures that follow are simple local regression (lowess) plots of the share of the vote for a given party and the share for a given electoral system in New Zealand’s general election and referendum of November, 2011. ((All percentages (really, shares) of votes in Part B of the referendum are calculated with the denominator being TOTAL votes in the referendum (i.e. informal votes are included).)) The data points themselves are suppressed because with around 6,000 polling places, the graphs would be overly cluttered. (For reasons indicated in a comment referenced below, the analysis is based on 5,375 polling places.)

Before I go any farther with this, I want to shout out a big THANK YOU to Vasi, who extracted the party-vote data from the many separate files posted at the Electoral Commission website.

Figure 1 shows the relationship between the support for each of New Zealand’s four largest parties–National, Labour, Green, and New Zealand First–and support for keeping the current MMP system. What stands out most is how linear the decline in support for MMP is as National support increases. Relative to Labour strength, support for MMP reaches a point where it levels off, but that is mainly because even where Labour receives around 50% support, the vote for MMP is already approaching 80%. Strong Green support is correlated with strong MMP support. No surprise there. There is a rather more limited relationship to votes for New Zealand First.

NZ11 figs 1&2

Figure 2 shows support for the two alternatives that were most likely to be favored by voters who wanted to change from MMP: Supplementary Member (SM, which is what I would call MMM) and First Past the Post (FP(T)P). Clearly, Labour strongholds were duly skeptical of SM, while support for both systems is associated with National’s strong polling places. There is actually a steady 30% or so support (on average) for FPTP in places where Labour wins more than about a third of the vote. I am not sure what to make of this; maybe strong Labour areas include many voters who long for the old days of Labour majorities under the former electoral system.

Figure 3 looks at the two large parties and support for the two options in Part B of the referendum that entail ranked-choice ballots: Preferential Vote (PV, which is usually known as the Alternative Vote) and the Single Transferable Vote (STV). Note that the lowess curves never rise above about 12% for either system. The one thing that stands out for me here is that there is actually some growing support for PV/AV as Labour strength increases. That makes sense, given that Labour could expect to benefit from many Green voters’ transfers. The sharp decline for PV/AV as National support increases is the mirror image of Labour’s trend–and is actually stronger, perhaps because National has so few potential allies who would give it second (or third, etc.) preferences. Still, we can’t make too much out of it, given that support for this system is so low overall. Not surprisingly, support for STV, which is nowhere high, decreases with support for either major party.

NZ11 figs 3&4

Figure 4 shows a rather interesting divergence of support for the two ranked-choice systems according to the strength of three smaller parties, Greens, New Zealand First, and United Future. Finally, we find out who wanted STV: Green voters! Well, actually, even they don’t want it very much, as places where their party vote reaches 40% or more still vote less than 20% for STV as their preferred alternative. Nonetheless, a trend for more STV support in locations with many Green voters certainly makes sense, inasmuch as STV is the only other proportional system on offer. It is a clear second best for supporters of a party like the Greens, which has generally dispersed support, but also (as the graph clearly shows) some pockets where it is locally quite strong. Local strength might enable it to combine with other parties and elect MPs in a few districts, while elsewhere its votes would transfer to Labour and help it elect an additional MP or two.

There is not much trend for the other parties, or for the Alternative Vote (PV), although places where there is a significant minority of United Future voters clearly are not places where voters see much benefit in the Alternative Vote.

Fun with Stata! And thanks again, Vasi, for your assistance!
See comment #6 for some information on the data.

10 thoughts on “New Zealand 2011: Party and electoral-system choice

  1. Glad to be of service! I don’t know if it’s useful to anyone, but I’ve thrown together the candidate votes per polling place, by party. It’s at the same location:

    I’m not sure what the merge failures are about, exactly. Some of them could be related to minutia like trailing whitespace, I don’t know how good Stata is with those issues. But there are at least a couple of data points I’ve checked and found in only one of the two data sets.

    Also a potential concern, the vote counts from each polling place in the referendum and the election aren’t the same. If there weren’t two separate ballots for the referendum and the election, this is a problem.

    • I am going to have to parse the merge a bit more closely whenever I have a chance. So the above comes with the caveat that some errors might be lurking, though I wouldnt expect the patterns reported to be changed in any big way.

      Stata can have issues with white spaces.

      And I am really glad you did the candidate votes, too! (Oh no, more merging!)

  2. The referendum and election ballots were on separate pieces of paper. From memory, there were separate ballot boxes to deposit them in (to assist with sorting, I assume). You can see the actual referendum voting paper on

    • Yes, there were also separate ballot boxes for the general election and referendum ballots. (I was able to visit a polling place as part of a program for visitors arranged by the Electoral Commission.)

      The general election ballot boxes were opened in the polling place immediately upon the close of the polls. The referendum ballot boxes were sealed and transferred to the Electoral Commission to be opened and counted at a later date. This is why the referendum results were released some weeks after the result of the general election was known.

  3. Presumably some election papers were in the referendum boxes. I assume they were extracted as part of the election final count.
    I note that there are significant numbers of advance and special votes, but the numbers involved in each district aren’t much larger than twice big polling booths.

  4. The problem with the data in the original post has now been solved–mostly. There is still the puzzle that the two data sets (referendum votes and party votes) have a different number of polling places. The difference is around 400 (out of just over 6,000 in each data set).

    I was able to get Stata to match up many more than it initially was able to do, by doing some searches and replace on annoying little differences between polling-place names in the two data sets. For instance, many Maori place names have accents or other marks that were rendered differently in the file-conversion process across the two data sets. And then there was the odd stray blank space, and that sort of thing. I gave up trying to match up the others where I could not see any obvious pattern to how they were rendered differently. So we are left with 691 polling places in the referendum data set that do not have corresponding data for parties, and 831 with the reverse. Some of these actually do not exist in the other data set. Others appear to be there, and my eye has failed to detect what character is making Stata think they are different.

    In any case, there appear to be no matches that are not real matches (as there were in the original that I posted), and it is unlikely that the un-matched polling places would change the patterns reported here. There does not appear to be any systematic pattern to the missing polling places (such as their being concentrated in a few electorates, etc.).

  5. There are a lot of artificial “polling places” such as “Auckland Central, Special Votes BEFORE polling day” that are only in one set or the other.

    There are also some polling place names that appear to be used in multiple electorates. I guess that’s because of the Maori electorates.

    So I tried matching polling places like this:

    1. Create unique names by concatenating the electorate name and polling place name, separated by a semicolon.

    2. Filter out artificial polls containing strings like “Special votes” or “Votes allowed” or “Polling places”.

    3. Normalize the names by lower-casing them and removing all characters except for letters, numbers, commas and semicolons. There don’t seem to be any cases of different polls being normalized to the same name, so this should be safe.

    I’m left with just 90 polls only present in party_vote, 11 only present in referendum_ppresults, and 5767 matches.

    • Vasi, that seems like a good procedure, and one I would have to follow if I were ever to turn this into an academic paper, Not sure I will ever get around to it, though.

Leave a Reply

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

You are commenting using your 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 )

Google+ photo

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

Connecting to %s