Tim Southee, Mike Hesson and the Sunk Cost Fallacy

Matt Henry has a better ODI strike rate with the ball than even Shane Bond, but can’t make the current Black Caps ODI side over players with much worse numbers

A sunk cost is an economic concept that refers to an expense that has been already paid for, so that this expense is irrecoverable. It sounds unremarkable, but sunk costs do funny things to the human brain. The Sunk Cost Fallacy, and a little game theory, may help explain why Mike Hesson refuses to make the hard call and drop Tim Southee for Matt Henry in the Black Caps ODI playing XI.

The Sunk Cost Fallacy is an example of a reasoning error that is commonly made when sunk costs are involved. It refers to when people do something irrational because they have sunk costs (in the form of money, time or energy) into a line of reasoning already.

The common example given is the Concorde project, during which the French and British governments realised that the project would never make economic sense, but which they continued with anyway on the grounds that they didn’t want to waste their sunk cost.

This is a well-known phenomenon in economics because it often leads to horrific waste, particularly when people throw good money after bad in the hope that their initial investment will be recouped (it’s also a well-known phenomenon in poker when players go broke). As far as Mike Hesson is concerned, the same psychological process may be occurring with his obstinate refusal to elevate Matt Henry to the opening bowler’s position alongside Trent Boult.

Southee is only 29 years old, but he has been playing for ages. He debuted as a teenager, and looked incredibly promising with a five-wicket haul and a run-a-ball 77 in a Test against England. Since then, New Zealand Cricket have invested substantial resources in him, giving him every opportunity to take the new ball against all comers. No less an authority than Allen Donald touted Southee as potentially a great swing bowler, but it’s hard to deny the raw numbers.

The numbers argue confidently that Southee is not as good as Henry (in ODIs at least), and probably never has been.

Since the loss to Australia in the final of the 2015 Cricket World Cup, Southee has averaged 42.77 with the ball. He has taken 45 wickets in this time frame at an economy rate of 5.72. Henry has taken 44 wickets at an economy rate of 5.81, which is similar, but has an average of 28.13 thanks to a strike rate of 29.

This is, amazingly, a better strike rate than Shane Bond managed over his career (29.2), and means that Henry has been 50% more likely to take a wicket on any given ball than Southee during this time.

Southee is striking at 44.8 since the last Cricket World Cup, and has not taken four wickets in an innings since then, despite playing in 38 games. Henry has only played in 25 matches since the final loss but has already managed three four-wicket bags and one five-wicket bag in that time – only one of each fewer than Southee has managed in a 132-match career.

In fact, Southee has not managed to get four wickets in an innings one time in the last three years, whereas Henry took four wickets in his last match.

Mike Hesson might be thinking here in terms of potential, in that, theoretically, Southee has the potential to be another Jimmy Anderson. Like Southee, Anderson is also tall, bowls with an open stance and relies on swinging the ball to nick batsmen out. Also like Southee, the Englishman didn’t achieve anything particularly special in his first 132 ODIs, returning an average of 30.18 for his 179 wickets.

But in 43 matches since the start of 2012, Anderson has taken 65 wickets at an average of 23.97. Hesson might be expecting a similar transformation to come over Southee, but it’s also very possible that he has invested so much time and energy in Southee that he sees this investment as a sunk cost that he is compelled to recoup.

A solution that would save everyone’s face would be to demote Southee to third seamer. This would be the best of all worlds for everyone except for Lockie Ferguson, who would then struggle for a starting berth.

The positives are that it would allow Boult and Henry, with the best strike rates, to bowl with the new ball when they are the most effective, and it would allow Southee to utilise his skill set of varied deliveries at the death while minimising his weaknesses of being slow and inaccurate. Southee also has 33 wickets as third seamer, averaging an entirely acceptable 28.18 (compared to over 35 while opening). It seems like a solution whose time has come.

*

If you enjoyed reading this essay, you can get a compilation of the Best VJMP Essays and Articles of 2017 from Amazon for Kindle or Amazon for CreateSpace (for international readers), or TradeMe (for Kiwis).

Understanding New Zealand: Turnout Rate in 2017

The turnout rate in the 2014 General Election was 77.9%, and at the time of the 2017 General Election this had climbed to 79.3%. This is not a huge change, but care must be taken not to be misled. Just because the overall turnout rate was about the same does not mean that the turnout rates of the various demographics within New Zealand society all remained the same. This article examines the deeper trends.

The most striking thing about the turnout rates in the 2017 General Election is that, despite the alarm about the huge numbers of immigrants New Zealand has absorbed in recent years, the election marked a sharp increase in turnout rate among the New Zealand-born.

The correlation between being New Zealand-born and turnout rate became much more positive from 2014 to 2017, from a significantly negative -0.24 to -0.10. This is arguably the story of the election and explains how we ended up with a nationalist party holding the balance of power.

Among the four major parties, the correlation with turnout rate and voting for a particular party remained very similar from 2014 to 2017 for National (0.76 to 0.75), Labour (-0.70 to -0.72) and the Greens (0.28 to 0.27). These slight falls were balanced by a fairly strong increase for New Zealand First (-0.09 to -0.02).

This doesn’t necessarily mean that there was a swing towards New Zealand First – it just means that the sort of person who is a New Zealand First supporter was more likely to vote this time around. Who they actually voted for requires further analysis.

If we look at the ethnic demographics, we can see that the correlation between being of a certain race and turnout rate between 2014 and 2017 strengthened for Kiwis of European descent (from 0.71 to 0.81) and for Maoris (from -0.75 to -0.68). These are the two ethnic groups most likely to support New Zealand First.

Pacific Islanders and Asians, who were more likely to be born overseas, were less likely to turn out to vote. The correlation between being a Pacific Islander and turnout rate was -0.58 in 2017, out from -0.44 in 2014, which makes them now almost as disenfranchised as Maoris. The correlation between being Asian and turnout rate was -0.22 in 2017, out from -0.10.

One reason for this is that even though large numbers of immigrants have turned up in New Zealand recently, many of these newcomers don’t seem to feel much of a connection with the country and so are not motivated to vote.

Where it gets complicated is that the correlation between median age and turnout rate increased from 2014 to 2017, from 0.77 to 0.79. This means that the people who voted this time around were older, and older people tend to vote National – but these voters did not vote National.

The correlation between being aged 20-29 and turnout rate became a lot more negative from 2014 to 2017, from -0.21 to -0.26. The correlation between being aged 30-49 and turnout rate followed a similar pattern, weakening from 0.21 to 0.13. The predictable result of this is older people voting more, and indeed we can see that the correlation between being aged 50-64 and turnout rate increased from 0.70 in 2014 to 0.73 in 2017, while the correlation between being aged 65+ and turnout rate increased from 0.64 in 2014 to 0.67 in 2017.

Unsurprisingly, then, the correlation between being on the pension and turnout rate increased from 2014 to 2017 (from 0.50 to 0.56). In fact, all of the benefit types apart from the student allowance also strengthened. The correlation between being on the invalid’s benefit and turnout rate strengthened from -0.53 in 2014 to -0.43 in 2017, and the correlation between being on the unemployment benefit and turnout rate strengthened from -0.76 in 2014 to -0.72 in 2017.

This is supported by the fact that voters were more likely to be New Zealand-born in 2017, because there is a significant correlation between being New Zealand-born and being on a benefit.

More information comes from noting that several correlations between belonging to privileged demographic categories and turnout rate decreased from 2014 to 2017. This applied to people working in information media and telecommunications (0.06 to -0.01), financial and insurance services (0.08 to 0.01) and professional and scientific services (0.28 to 0.23). Also, the correlation between having never smoked tobacco and turnout rate fell from 0.35 in 2014 to 0.25 in 2017.

On the other hand, the correlations between more working-class occupations and turnout rate increased, most strikingly so in the occupations that involved the most personal contact. The correlation between working in a particular occupation and turnout rate increased from 2014 to 2017 in the case of education and training (-0.10 to -0.03), healthcare and social assistance (-0.04 to 0.05), arts and recreation services (0.04 to 0.09) and hospitality (-0.09 to -0.01).

A poorer cross-section of the population turned out to vote in 2017, which is another clue as to where Labour won. All of the correlations between being in an income band below $70K and turnout rate strengthened from 2014 to 2017, and all of the correlations between being in an income band above $100K and turnout rate weakened from 2014 to 2017. The correlation between being in the $70-100K income bracket and turnout rate remained exactly the same, at 0.38.

Another striking correlation is that between being part-time employed and turnout rate, which rose sharply from 0.45 to 0.58 between 2014 and 2017. This, coupled with what we know about income brackets and turnout rate, suggests that it was the people on the margins between doing well and doing poorly who shifted from National to Labour. It may be that these people saw the promise of the country being lost, or felt that they missed out on all the loot of the last nine years.

Perhaps the clearest sign of where National lost the election comes from the correlations with the flag referendum. The correlation between voting National in 2017 and voting to change the flag in the second flag referendum was an extremely strong 0.93, which tells us that it was pretty much only National voters to wanted to change the flag to the National Party version.

These National Party supporters, being generally well enfranchised, have very high turnout rates. The correlation between turnout rate in 2014 and voting to change the flag was 0.75, exactly the same as the one between turnout rate in 2017 and voting to change the flag. However, the correlation between turnout rate in the second flag referendum and turnout rate in the general election increased from 2014 (0.86) to 2017 (0.92).

This suggests that many of the new people who voted in the 2017 General Election but did not vote in the 2014 one were Labour supporters who came from generally National-supporting demographics (i.e. wealthy but not too wealthy, old but not quite a pensioner, white, employed and part-time employed, male). Had they been National supporters, the correlation between turnout rate and voting to change the flag would have increased from 2014 to 2017, because the vast bulk of people who wanted to change the flag were National supporters.

We can say that it was here that the centre, and thereby the 2017 General Election, was lost by the National Party.

*

The second edition of Understanding New Zealand, by Dan McGlashan, was published by VJM Publishing in the summer of 2017/18. It contains all the analysis of Kiwi voting patterns and demographics you could ever want!

The 10 Most Uncorrelated Pairs of Things in New Zealand Society

The third part of Dan McGlashan’s Understanding New Zealand contains an index of all the 11,628 correlations considered in the writing of the first two parts of the book. There are so many of these that the index has to go to three decimal places in order to properly distinguish them from each other. Even then, there are some pairs of variables that are so perfectly uncorrelated that they go down to weaker than 0.000 – as this article examines.

This data was collected from the Electorate Profiles that can be found on the Parliamentary Library website, and then entered into a Statistica database from which a correlation matrix was calculated. These ten pairs of variables represent the least correlated things in New Zealand society.

9=. 0.0007, Voting New Zealand First in 2014 and Being of European descent

New Zealand First gets much more support from Maori electorates than most people realise. Their level of support among Maoris declined in 2017 but in 2014 it was strong enough that there was no correlation between voting New Zealand First and being white.

9=. -0.0007, Voting Internet MANA in 2014 and Having an income between $60-70K

Internet MANA voters were a fair bit poorer than average, and so there is no correlation between voting for them in 2014 and having a slightly higher than average income.

8. -0.0006, Having NZQA Level 2 as a highest academic qualification and Working as a sales worker

Leaving school at the end of Sixth Form is a middling sort of academic achievement, and being a sales worker is an extremely common line of employment.

7. 0.0005, Being a Brethren and Working in public administration and safety

Brethrens are indifferent to working for the Government.

6. -0.0004, Voting Internet Party in 2017 and Being a Spiritualist or New Ager

Internet Party voters are young and mostly live in Auckland; Spiritualists and New Agers are a bit older and are distributed reasonably evenly throughout the country.

3=. 0.0003, Voting to change the flag in the second flag referendum and Being in a couple with children

Couples with children probably have more important things to think about than a flag referendum, which was only ever a vanity project for the elderly elite with leisure time.

3=. -0.0003, Being a Christian (not further defined) and Being a sales worker

Christians, when not further defined, are generally a bit poorer than average, and sales workers are moderately wealthy, hence no correlation.

3=. -0.0003, Being a Pacific Islander and Being on the invalid’s benefit

Pacific Islanders are stereotyped as being on the benefit a lot, but the truth is that the sort of Pacific Islander who wanted to go on the invalid’s benefit probably wouldn’t be motivated to immigrate to New Zealand in the first place.

1=. -0.0001, Voting Internet MANA in 2014 and Taking a bus to work

Internet MANA voters in 2014 tended to live in the Far North, and those who lived in Auckland often took a bus to work, the others not so much.

1=. -0.0001, Being a regular tobacco smoker and Being a sales worker

There are no more middle-of-the-road variables than being a regular tobacco smoker and being a sales worker. Both traits point to a very normal, standard sort of person, hence no correlation.

The Boringest T20I Is Worse Than The Boringest ODI

Scoreboard pressure in often so intense in T20s, especially at international level, that one misstep puts the chasing side in a hopeless position

In much the same way that cricket analysts were slow to catch on to the importance of strike rate in ODIs, so too have they been slow to catch on to the importance of strike rate in T20s. Everyone knows strike rate is important in limited overs; what few understand is how this can lead to extremely boring matches. This article looks at why efforts to contrive a more interesting game of cricket have only somewhat succeeded.

The advent of T20 came about when someone realised that, all other things being equal, fans liked seeing boundaries and wickets, and didn’t much care for dot balls or for contests that were over long before they technically finished because the chasing side lost early wickets. And so a form of the game was contrived to have as few dot balls as possible, and as many boundaries and wickets as possible.

But what this new fashion risks losing sight of is the fact that cricket is only interesting in the first place because it is a contest of skill, and sometimes the nature of the match situation in T20 is not conducive to playing the game skillfully.

For instance, it almost never occurs in ODIs that the chasing team, in the first five overs, falls so far behind the required strike rate that the match is effectively lost without resorting to slogging. The chasing team might lose early wickets, which makes the chase much harder, but as long as they can keep their wickets intact there are still plenty of overs in which to win through playing proper cricket and building an innings.

In T20s it’s common for the chasing team to build up enough scoreboard pressure in the first five overs that they effectively cannot win. In other words, it’s possible to lose the game with the bat in the first five overs because of scoreboard pressure – something that is near to impossible in ODIs.

This can happen if the team setting a total bats well enough that they’re close to the optimal possible run rate over the 20 overs (which is usually somewhere just above 10 an over, a rate that cannot realistically be maintained for an ODI). When the required run rate for the batting team climbs above this, it cannot be achieved without taking risks (i.e. slogging) and when this happens one is no longer playing cricket. It’s no longer a contest of skill but merely hit and hope.

This is an incredibly boring outcome from the spectator’s point of view, because the beauty of cricket is that it is a contest of skill where the batsman must find a balance between aggression and keeping his wicket intact. Removing the “keeping the wicket intact” part of the equation often reduces the game to slogging.

It has happened twice in the two matches of the T20I series between New Zealand and India so far.

In the first T20I in Delhi this week, Shikhar Dhawan and Rohit Sharma both scored 80 at strike rates of 145+ opening the India innings, which allowed them to finish on 202, more than 10 runs an over. This asking rate was so high that the slightest stumble in the chase would see the match over as a contest.

And it duly was effectively over, 22 balls into the chase after Munro and Guptill had both been dismissed. At this point the equation was 185 runs off 98 balls and Kane Williamson and Tom Latham, for all their undoubted skill as batsmen, couldn’t do much about it because neither could slog at the required rate.

Once a T20 match gets to this stage it gets ugly, because batsmen have no time to get used to the conditions so as to play big innings where they strike the ball skillfully and with timing into gaps in the field. So after Williamson and Latham were dismissed, Colin de Grandhomme had to come in and slog his first ball, which got him out.

In last night’s T20I at Rajkot, the contest was effectively over in the second over of the chase after Trent Boult had taken two wickets. Colin Munro clubbed his way to 109* (58) in the first innings, which meant that the slightest stumble would put the total out of reach.

And it duly was effectively over after Boult’s second, which meant they needed 10.5 an over for the remaining 18, or an individual strike rate of 175, with their two best hitters out. None of their batsman managed this, as few do.

The remaining batsmen were unable to score at the required strike rate without throwing their wickets away, and so limped to a loss, effectively throwing in the towel long before the innings concluded.

In ODIs strike rate is a major factor, but strike rate seldom destroys the chasing team’s hopes as quickly, as ruthlessly and as completely as in T20Is. Scoreboard pressure can destroy a chasing team’s chances so quickly that players can be forced into mindless slogging, and this is much less interesting to watch than a balanced contest between bat and ball.

The Black Caps T20 side already can’t find room for batsmen of the quality of Ross Taylor and Tom Latham, because they can’t (or don’t) slog enough. This means that the viewer is watching batting of a lower level of skill just because the way the T20 format is contrived promotes such. This is also not interesting from a viewer’s perspective.

The real horror scenario is this – what if it is decided that strike rate is so important in T20Is that there’s no room in the side for Kane Williamson? Because if this day ever comes, T20Is might start to be considered a joke format by cricket fans in the same way most rugby union fans consider sevens, or at least one only suited for levels below international.

If the joy of sport comes from watching skill on display then it can be argued that T20Is are objectively more boring than ODIs because they objectively give less opportunity for the batsman to display skill. The overwhelming importance of strike rate means that skillful cricketers are often pushed out of the side by sloggers, which are only fun in small amounts.

New Zealand in India ODI Series 2017, First ODI Preview

Colin Munro’s T20 form has seen him elevated to open for the ODI side in the hope that he can replicate Brendon McCullum

A lot of things are going against the Black Caps before their One Day International series against India in India starts tonight (Sunday) at 9p.m. NZT. Ominously, India sit atop the ICC ODI team rankings table, equal to South Africa with 120 ratings points, while the Black Caps are currently a mid-table side at 5th. Worse, the matches are in India, where India just demolished the world champion Australia side 4-1.

On the other hand, a lot is going in their favour. New Zealand scored 343/9 in their last warm-up match against an Indian Board President’s XI, with centuries to Ross Taylor and Tom Latham, who in all likelihood will comprise the 4-5 axis over the series. Considering that Kane Williamson will bat at three, this suggests that India might have difficulty bowling the Black Caps out.

Colin Munro is expected to open the batting with Martin Guptill, which is an experimental measure intended to fill the gap left by Brendon McCullum at the top of the order. McCullum’s explosive starts made it possible for Williamson and Taylor to build innings without risk, and Munro has a T20 strike rate of close to 150 – close to what McCullum was striking at for the last 2 years of his ODI career.

It looks as though Williamson will try and get 10 overs out of Colin de Grandhomme, who will bat 7, and with Latham at 5 doing the wicketkeeping this leaves a spot for a pure batsman at 6. This spot might get filled by Henry Nicholls, as it was during the final warm-up match, or the Black Caps might start playing hitters from there, which would probably mean Glenn Phillips.

The other question mark is whether or not Tim Southee is still good enough to command a starting spot in this ODI squad. Although he has been a first-choice seamer for a handful of years, his ODI bowling average over the past three years – which includes the great run at the 2015 Cricket World Cup – is 36.47, which is only good enough for 33rd place on the official ODI bowling rankings.

Damningly, there are four players from Afghanistan alone higher than Southee on the official rankings, which is probably not good enough for a Black Caps side that made the final of the last Cricket World Cup and is aiming to go one better in England in 2019.

It’s very possible that the Black Caps play two spinners on the slow Indian decks, which means they take both Mitchell Santner and Ish Sodhi. Assuming then that they still need de Grandhomme at 7 for the sake of the batting, and that Trent Boult is undroppable, that means Southee will be competing with Adam Milne and Matt Henry for that second seamer’s position.

India, for their part, have two extremely crafty bowlers. With the new ball is Jasprit Bumrah, who isn’t quick but has a Nathan Bracken-style range of subtle variations that make him hard to hit off the square, and with the older ball is Axar Patel, whose wily orthodox is also a difficult proposition on Indian surfaces.

The likely winning of the game, however, will be in India’s power-packed top order. They have four batsmen ranked in the top 14 in the world, with two of them – Virat Kohli and Rohit Sharma – ranked higher than Kane Williamson. Kohli averages a monstrous 55.13 from almost 200 matches, and getting rid of him early is simply a necessity if the Black Caps are to win.

Alongside them are the silky skillful Ajinkya Rahane and the brutal Shikhar Dhawan at the top and MS Dhoni with his 50+ average down the order. India have also found a genuine allrounder in Hardik Pandya so have no obvious weaknesses anywhere.

The Black Caps are only playing three ODIs and they are paying $3.80 on BetFair to win the first one, compared to India’s $1.35. This suggests that the market is expecting a 2-1 or 3-0 win to India.

Who Voted for the Ban 1080 Party?

Of all the smaller parties in the 2017 election, the Ban 1080 Party might be the strangest of them. There are other small single-issue parties – the Aotearoa Legalise Cannabis Party being foremost of these – but even these other parties have equivalents overseas. Who are the Ban 1080 Party, and what do we know about their 3,005 voters?

The Ban 1080 Party website argues for the need to stop making aerial poison drops that use sodium monofluoroacetate (1080) in New Zealand’s national parks and forests. The website’s tagline is “Protect our native birds” and they believe that aerial 1080 drops are a risk to the wellbeing of New Zealand’s birdlife.

A strong South Island focus was evident from the correlation matrix – the correlation between voting Ban 1080 Party in 2017 and living on the South Island was 0.37. The reason for this is probably because a greater proportion of South Islanders will live in the vicinity of a national park or a forest than North Islanders, who are much more urban on the whole, and it’s these people who access the outdoors who are most concerned about things like aerial poison drops.

This explain why the Ban 1080 Party also correlates strongly with other demographics that are well-represented on the South Island. The correlations between voting Ban 1080 in 2017 and other demographic categories were 0.34 for being a Kiwi of European descent, and 0.22 with median age. The only age bracket with a significant positive correlation with voting for the Ban 1080 Party was the 50-64 age bracket – the correlation here was 0.38.

If we examine measures of class we can see that Ban 1080 Party voters are poorer and less educated than the national average, which is especially striking if one considers that they otherwise belong to demographics that are positively correlated with wealth.

The correlation between voting Ban 1080 Party in 2017 and median personal income was -0.23, and the correlation between voting Ban 1080 Party in 2017 and having no NZQA qualifications was 0.44. Related to this is a correlation of 0.30 between voting Ban 1080 Party in 2017 and being a regular tobacco smoker. This paints a picture of a section of the community who are relatively simple people and who perhaps have been taken in by the hysteria a bit.

The rural nature of Ban 1080 Party voters is demonstrated starkly when it comes to the correlations between voting for them in 2017 and working in the agriculture, fishing and forestry (0.67) or mining (0.69) industries. There was also a significant positive correlation of 0.35 between voting for Ban 1080 Party in 2017 and working in the hospitality industry.

These three correlations reflect the high proportion of Ban 1080 Party voters who were enrolled in either the West Coast-Tasman or Clutha-Southland electorates.

Underlying that Ban 1080 Party voters are comprised of the outdoorsy kind of person who spends a lot of time in national parks and forests, there are significant positive correlations between voting Ban 1080 Party in 2017 and both being born in New Zealand (0.46) and being male (0.35).

Ironically, given their heavy conservation focus, the Ban 1080 Party does not attract followers who are like the Green followers. The correlation between voting Ban 1080 Party in 2017 and voting Green in 2017 was -0.09.

People who voted Ban 1080 Party tended to overlap with those who voted New Zealand First and, oddly, Aotearoa Legalise Cannabis Party. Voting for the Ban 1080 Party in 2017 had a positive correlation with voting for either of these parties: 0.41 for New Zealand First and 0.28 for the ALCP.

Part of the reason for this is the high level of Maori support for the party. This might sound contradictory, given that most Maoris live on the North Island, but a couple of statistics make this association clear. The first is the correlation between voting Ban 1080 Party and being Maori, which was 0.16, and the correlation between living on the South Island and being Maori, which was -0.26.

This tells us that South Island Maori were proportionately big supporters of the Ban 1080 Party, which is fitting considering that this demographic is extremely active in the outdoors with hunting and food gathering.

Who Voted New Zealand First in 2017

New Zealand First voters are generally drawn from the hard-done-by segments of the population

A previous article in this column examined the differences between New Zealand First voters and the voters of both National and Labour. It turns out that New Zealand First is almost equidistant from the two major parties if measured demographically. This article, however, looks more closely at who voted for New Zealand First in particular.

Despite being considerably whiter than it was in 2014, New Zealand First is not a particularly white party. In 2014 there was a correlation of 0.00 between voting New Zealand First and being a Kiwi of European descent. By 2017 this had climbed to 0.21, which was still not significant.

The strongest correlation between being of a particular ethnicity and voting New Zealand First was with being Maori, which was 0.38. The correlations between voting New Zealand First in 2017 and being a Pacific Islander (-0.21) or being Asian (-0.52) were much more negative.

Tellingly, for a nationalist party, their strongest support was from Maoris, who have the strongest roots in the country as essentially none of them are immigrants. Their next strongest level of support was from Kiwis of European descent, who have the second-deepest roots in the country, and their weakest level of support was from Asians, who have the shallowest.

As in 2014, New Zealand First voters in 2017 were some of the least educated out of any voting bloc. The correlation between having no NZQA qualifications and voting New Zealand First in 2017 was 0.67, and the correlation between having a doctorate and voting New Zealand First in 2017 was -0.60. This tells us that New Zealand First voters are decidedly working-class.

True to stereotype, there was a strong positive correlation of 0.58 between voting New Zealand First in 2017 and being on the pension, but there was also a strong positive correlation of 0.47 between voting New Zealand First in 2017 and being on the invalid’s benefit.

All of this suggests that the easy story of New Zealand First being an old racist’s party is somewhat misguided – it’s true that they do get many votes from poorly educated old white people, but that’s more because New Zealand First gets a lot of votes from hard-done-by people in general and poorly educated old people tend to be limited to their pension and therefore hard-done-by.

Gareth Morgan’s personal antipathy towards Winston Peters was reflected in the correlation of -0.31 between voting The Opportunities Party in 2017 and voting New Zealand First in 2017. This put TOP in a band with United Future (-0.27), ACT (-0.34) and the Greens (-0.48) as parties whose voters did not correlate highly in a general demographic sense with the voters of New Zealand First.

All four of those parties are particularly Pakeha-heavy parties, and ACT, the Greens and TOP appealed heavily both to young and educated people. So there is plenty of reason for these reasonably strong negative correlations.

There were positive correlations between voting New Zealand First in 2017 and voting for any of the Maori-heavy parties in 2017. These were Maori Party (0.11), MANA (0.24) and Aotearoa Legalise Cannabis Party (0.40).

Many will be surprised that there is a moderately strong positive correlation between voting New Zealand First in 2017 and voting Conservative Party in 2017 – this was 0.38. The reason for this is that both parties appeal to the large faction of poorly-educated old white voters mentioned above.

Despite the shared appeal to old white people, however, the correlation between voting National in 2017 and voting New Zealand First in 2017 was only 0.04, far from being significant. The reason for this is the class difference – the National Party appeals to people who are doing well economically (and most of these people are old), whereas New Zealand First appeals to people at the bottom of the ladder (and poor old people with no realistic way of becoming wealthier are definitely near the bottom).

*

This article is an excerpt from the 2nd Edition of Understanding New Zealand, which Dan McGlashan and VJM Publishing will have ready for sale at the end of October 2017. This will contain statistics calculated according to the official final vote counts and will be freshly updated with data from the 2017 General Election.

Who Voted Greens in 2017

The Green Party tore itself to pieces during this year’s electoral campaign, but a hard core of voters stayed with the party

The Green Party vote collapsed from 2014, as a previous article has examined, with much of it going to The Opportunities Party. Although the special votes helped them out since the time that linked article was written, they still fell to 6.27% in 2017 from 10.70% in 2014. This article looks at who voted for them.

The major curiosity about the Greens and their movement is that, although they are on the left, they are comprised of people who do not immediately benefit from increased resource distribution (i.e. the wealthy). The correlation between voting Greens in 2017 and median personal income was 0.36, which was up from 0.31 in 2014, and not a whole lot weaker than the correlation of 0.49 between median personal income and voting National in 2017.

All of the correlations between voting Green and being in one of the income bands below $70K were weak no matter if they were positive or negative. But above this point, the correlations were strong. Between voting Green in 2017 and earning $70-100K the correlation was 0.49, with earning $100-150K it was 0.56, and with earning $150K+ it was 0.51.

However, much like 2014, the average Green voter in 2017 was a bit younger than the average Kiwi. The correlation between voting Green in 2017 and median age was -0.18. This is mostly because of a very strong correlation of 0.60 between being aged 20-29 and voting Green in 2017.

The Greens lost ground with Kiwis of European descent. By 2017 the correlation between voting Green and being a Kiwi of European descent was 0.17, down from 0.24 in 2014, which meant that although it was still positive it was no longer significantly so. They also lost ground with Maoris. The correlation between being Maori and voting Green was -0.09 in 2014 but -0.14 by 2017.

By 2014, the Greens were already much better educated than the average Kiwi, and by 2017 this distinction had only strengthened. The correlation between voting Green in 2017 and having a university degree was 0.64 for having a Bachelor’s, 0.81 for having an Honours, 0.71 for having a Master’s and 0.68 for having a doctorate. This higher general education explains why Green voters can be above average in income despite being below average in age.

The Greens maintained their core, urban elite vote despite the losses from 2014, and this is evident from looking at the voting patterns of certain industries. The correlation between voting Green and working in information media and telecommunications was 0.75, with working in professional and scientific services it was 0.70, and with working in arts and recreation services it was 0.69. All three of those correlations were as strong or stronger in 2017 than they were in 2014.

Things were much different for voters in working-class industries. In 2017 the negative correlations between voting Green and working in a particular industry included -0.02 in retail trade (down from 0.09), -0.20 in construction (down from -0.09), -0.29 in agriculture, forestry and fishing (down from -0.24), -0.32 in transport, postal and warehousing (down from -0.29) and -0.56 in manufacturing (down from 0.49).

A couple of correlations that Green Party thinkers won’t be at all happy about, given their pretensions to being a party that represents the poor and downtrodden, are the moderate negative ones between voting Green in 2017 and being a machinery operator and driver (-0.47), labourer (-0.31) or as a technician or trades workers (-0.25). These occupations are dominated by Maoris who tend to have pro-Labour and pro-New Zealand First sentiments.

Green voters had little in common with the voters of any other party except for The Opportunities Party. Lending further evidence to the suggestion that TOP primarily took votes away from the Greens is the fact that the correlation between voting TOP in 2017 and voting Greens in 2017 was 0.77.

None of the correlations between voting Green in 2017 and voting for the parties that did get into Parliament were significant, except for the case of New Zealand First, which was significantly negative. These were 0.17 with ACT, 0.11 with Labour, -0.25 with National and -0.48 with New Zealand First.

It might seem strange that Green Party voters have a stronger correlation with ACT Party voters than with Labour Party ones. That’s not really so strange if one considers that on measures such as age, education, income and ethnicity, the two parties are reasonably similar (i.e. young, well-educated, rich, white and urban).

In a sense, it can be said that the Labour-National dichotomy is the dilemma the average Kiwi voter is faced with, but the ACT-Green dilemma is the one that the average ambitious, professional young Kiwi voter is faced with.

*

This article is an excerpt from the 2nd Edition of Understanding New Zealand, which Dan McGlashan and VJM Publishing will have ready for sale at the end of October 2017. This will contain statistics calculated according to the official final vote counts and will be freshly updated with data from the 2017 General Election.

Who Voted ACT in 2017

The ACT Party got an extremely high level of media coverage for a party so disliked by the electorate, but it didn’t help them in 2017

The ACT Party cuts a lonely figure on the New Zealand landscape. Although their advertising budget was literally hundreds of times greater than the Aotearoa Legalise Cannabis Party, they couldn’t win even twice as many votes. This tells a story of a party whose goals are not well aligned with the will of the New Zealand people; some could argue they were directly antithetical. This article looks at who their voters were.

ACT won 13,075 votes in 2017, down from 16,689 in 2014. This suggests that they were abandoned by some 20% of their voters from 2014. Fortunately, the correlation matrix tells us about several places that their vote became weaker.

ACT voters were the wealthiest of any party’s voters. The correlation between voting ACT in 2017 and median personal income was 0.61, which was a fair bit higher than the correlation of 0.36 between voting ACT in 2014 and median personal income (for reasons that we will investigate).

Not even the correlation between median personal income and voting National in 2017 was as strong – this was 0.49. In 2014 the average ACT voter was poorer than the average National voter, so the fact that they are wealthier in 2017 is a useful clue. It suggests that some ACT voters nearer the middle of the income ladder switched allegiances.

The correlations with voting for ACT and being in the higher income bands strengthened from 2014 to 2017. For those earning $150K+, the correlation with voting ACT in 2017 was 0.79, up from 0.44 in 2014. For those earning $100-150K, it increased to 0.66 in 2017 from 0.43 in 2014, and for those earning $70-100K it increased to 0.53 in 2017 from 0.35 in 2014.

At least part of the reason for these explanations is because the correlations between having a university degree and voting ACT strengthened from 2014 to 2017. By 2017 the correlation between voting ACT and having a university degree was 0.70 for having a Bachelor’s, 0.58 for having an Honours, 0.65 for having a Master’s and 0.51 for having a doctorate.

ACT also became a lot whiter from 2014 to 2017. By 2017 the correlation between voting ACT and being a Kiwi of European descent had become 0.16, much more positive than 2014 when it was -0.28. This reflects a collapse in Asian support – the correlation between being Asian and voting ACT was 0.85 in 2014, but only 0.46 by 2017.

These correlations start to tell a story of a large number of Asians who left the ACT Party for the National Party after 2014 (this is supported by the investigation into who voted for the National Party in 2017).

Despite the initial assumption made by many, the large numbers of Asians voting National could actually speak to an increasing solidarity between Asians and other Kiwis, because although National speaks for low taxes and low welfare they aren’t as aggressive about it as ACT are.

Pacific Islanders don’t like ACT (the correlation between being a Pacific Islander and voting ACT in 2017 was -0.23) and Maoris really don’t like ACT (the correlation between being Maori and voting ACT in 2017 was -0.51). This is not surprising if one considers how fervently ACT support the wealthy.

One curiosity is that the ACT voting bloc became older this election. The correlation between median age and voting ACT in 2017 was 0.26, compared to 0.02 in 2014. This is hinted at by the strengthening in the correlations between voting ACT and being aged 50-64 (from -0.07 in 2014 to 0.17 in 2017) and between voting ACT and being aged 65+ (from -0.11 in 2014 to 0.11 in 2017).

Considering also that the average ACT voter was much less likely to be born overseas in 2017 (the correlation between voting ACT and being born overseas fell from 0.78 in 2014 to 0.57 in 2017), this paints a picture of a rich, old, white, very highly-educated core of ACT voters who have remained with the party, and a less-committed group of younger, heavily Asian professionals, some of whom still supported ACT in large numbers in 2017, but many of whom were successfully tempted to switch allegiance to the National Party.

*

This article is an excerpt from the 2nd Edition of Understanding New Zealand, which Dan McGlashan and VJM Publishing will have ready for sale at the end of October 2017. This will contain statistics calculated according to the official final vote counts and will be freshly updated with data from the 2017 General Election.

Who Voted For Labour in 2017

Jacindamania might have won Labour an extra 10%, but these new voters were not representative of Labour voters as a whole

The Labour Party had a rocky ride leading up to the 2017 Election, with the resignation of then-leader Andrew Little forcing a rethink of their entire electoral campaign. Despite that, their vote increased to 36.89% after the special votes were all counted, up from a dismal 25.13% in 2014. A previous article has already covered who those new voters were – this one looks at the overall group of Labour voters as a whole.

Labour voters are much poorer than the average New Zealander. The correlation between voting Labour in 2017 and median personal income was -0.52, about the same as in 2014, and shows the degree to which Labour is in favour of greater resource distribution (or at least the degree to which it gets support from those in favour of such).

There were significant positive correlations between being in any income band below $15K and voting Labour in 2017. This signifies a large change among those in the $10-15K income band – being in this band had a correlation of 0.21 with voting Labour in 2014, and had increased to 0.35 by 2017. Also, being in the $30-35K income band had a correlation of 0.15 with voting Labour in 2014, increasing to 0.27 by 2017.

Being in any of the income bands above $60K was significantly negatively correlated with voting Labour in 2017. This tells us that the idea of redistributing wealth through taxation doesn’t appeal much to the sort of people who have the most wealth.

There are a number of reasons for this connection between voting Labour and relative poverty.

One of the most obvious reasons is that Labour voters are much younger. There is a very strong negative correlation of -0.75 between voting Labour in 2017 and median age. Although part of the reason for this is that children are included in the median age statistics and that parents of young children prefer to vote Labour, most of the reason is simply because young people are poorer.

Because we already know there is a significant correlation between age and wealth we can tell that much of the reason why Labour voters are younger and poorer is simply because they have had less time to build a career or gain job expertise and this lack of seniority results in lower wages and salaries, which results in relatively stronger sentiments in favour of wealth distribution.

Labour voters are also much less educated. The correlation between having no NZQA qualifications and voting Labour was 0.34 in 2014 and 0.38 in 2017. The correlations between having no NZQA qualifications and voting New Zealand First were 0.79 in 2014 and 0.67 in 2017, so this suggests that a large number of the working class shifted from New Zealand First to Labour in the three years before the 2017 Election.

The university educated are mildly unwilling to vote Labour. The correlations with voting Labour in 2017 and having a university degree were -0.24 for a Bachelor’s, -0.22 for an Honours, -0.19 for a Master’s, and -0.17 for a doctorate. These aren’t strong – only the first of them is even statistically significant – but they’re less strongly negative than in 2014.

This mild unwillingness is probably down to two contrasting factors. Most university educated people are young because tertiary education became liberalised in recent decades and available to many more people, but on the other hand university educated people earn a lot more money than their non-educated peers, and this improved social position inclines them away from policies of resource distribution.

A third factor explaining the correlation between poverty and voting Labour is that Labour voters are much more likely to be on a non-pension benefit. The correlations between voting Labour in 2017 and being on a benefit were 0.41 for the student allowance, 0.58 for the invalid’s benefit and 0.73 for the unemployment benefit. Related to this is the fact that being a solo parent had a correlation of 0.79 with voting Labour in 2017.

In 2017, the Labour vote correlated strongly with the votes for other parties who also have a high level of Maori support. Here the correlations with voting Labour were 0.61 for the Maori Party, 0.56 for the Aotearoa Legalise Cannabis Party and 0.41 for MANA.

Also in 2017, the Labour vote had strong negative correlations with the vote of parties full of old, white men who don’t want to redistribute resources. Most obviously was with National (-0.94) but significant negative correlations also existed with ACT (-0.62), United Future (-0.43) and the Conservatives (-0.31).

The Greens and The Opportunities Party were close to neutral in this regard. The correlation between voting Labour in 2017 and voting Greens that year was 0.11, whereas with voting TOP it was perfectly uncorrelated at 0.00.

The biggest change from 2014 was with the correlation between voting Labour and voting New Zealand First, which was 0.11 in 2014 but had become negative by 2017, at -0.15. The reason for this is mostly because of the large number of Maori voters who left New Zealand First after they said they wanted to abolish the Maori seats, and this can be seen by the change in correlations between being Maori and voting New Zealand First (down to 0.38 in 2017 from 0.66 in 2014) and between being Maori and voting Labour (up to 0.58 in 2017 from 0.42 in 2014).

Woman were significantly more likely to favour the Labour Party in 2017. The correlation between being female and voting Labour in 2017 was 0.33, slightly stronger than in 2014. As mentioned in the section about National, the reasons for this can be surmised from evolutionary psychology.

In summary, the sort of person who would vote Labour is someone in favour of greater resource distribution, which means someone with less resources than average, which means young people, the less educated, Maoris, Pacific Islanders and women.

*

This article is an excerpt from the 2nd Edition of Understanding New Zealand, which Dan McGlashan and VJM Publishing will have ready for sale at the end of October 2017. This will contain statistics calculated according to the official final vote counts and will be freshly updated with data from the 2017 General Election.