Saturday 31 December 2016

2016 in Review

The Ghost of Blogging Present

There was a lot that happened this year, and as a result I have not had time to do a full break down of our predictions and the like. Besides the obvious—the election of Donald Trump in November—this year saw an alteration to Australia's upper-house voting system, a federal election, Britain's decision to leave the EU, elections in the NT and ACT, several overseas elections mentioned in passing including Austria, the Philippines and Switzerland, and the US Primaries.

It is useful to consider the results, which I will be doing hastily before [edit: apparently also after] midnight here, but I also think this might be a useful tradition in future years as a part-recap as well as picking up the many balls I will inevitably drop during the course of the years to come.

The Ghost of Predictions Past

Of the elections mentioned above, four races were given concrete predictions on this blog: the United States Primaries, the Brexit vote, the Australian Federal Election and the US Presidential Race.

US Primaries

My earliest predictions on the Republican candidacy in the primaries were problematic in that they consistently gave Trump a clear win in every seat. While the momentum of Trump support was high, it is unlikely he would have won every state. However, Trump's last contestants dropped out shortly after so in a sense the results were accurate.

The real predictions, however, were around the Democratic primaries and largely based on racial analysis. Some states had already been called, of course, as these were the source of our base data. The predictions were as follows:

STATE PREDICTION RESULT
ALASKA SANDERS SANDERS
CALIFORNIA SANDERS CLINTON
CONNECTICUT SANDERS CLINTON
DELAWARE CLINTON CLINTON
DISTRICT OF COLUMBIA CLINTON CLINTON
HAWAII SANDERS SANDERS
KENTUCKY SANDERS CLINTON
MARYLAND CLINTON CLINTON
MISSOURI SANDERS CLINTON
MONTANA SANDERS SANDERS
NEW JERSEY CLINTON CLINTON
NEW YORK CLINTON CLINTON
OHIO CLINTON CLINTON
OREGON SANDERS SANDERS
PENNSYLVANIA SANDERS CLINTON
RHODE ISLAND SANDERS SANDERS
SOUTH DAKOTA SANDERS CLINTON
UTAH SANDERS SANDERS
WASHINGTON SANDERS SANDERS
WEST VIRGINIA SANDERS SANDERS
WISCONSIN SANDERS SANDERS
WYOMING SANDERS SANDERS

This resulted in 16 correct predictions our of 22, or around 73%. While not a spectacular result, this is better than the 50:50 of random guessing (~11 correct predictions expected), or just giving every state to Clinton or to Sanders based on who was expected to win the most states (with 12 and 10 correct predictions respectively).

The main limitation on refining this data in the future is the absence of reliable state-by-state polling.

Brexit

There isn't a huge amount to summarise here. There was a single prediction. It was wrong.

I could point to the number of other predictions and polls that also got this one wrong, the general consensus that the Leave vote couldn't win leading it's opponents to vote in a contrary manner in protest or else stay home out of apathy, followed by a period referred to as Bregret. And in some ways it is true that the polling generally showed a strong chance for the Remain camp. But the polling was far from unanimous and it was evident that the Leave vote (in blue below) was gaining as time went on not only in many separate polls but also in my own aggregation.

While it is often convenient to simply say “the polls were wrong” (I'm looking at you too, US presidential results), the fact is that I chose to rely on that flawed data without any attempt to correct it and, more problematically, when the data led to a conclusion I found unlikely I dismissed the data too readily as flawed by its collection method.

Australian Federal Election
Two sets of predictions were made here: one set for the lower house and one for the upper. The latter of these was complicated by the introduction of a new voting system with multiple above-the-line votes and exhaustion introduced. This replaced the preference tickets that offered some reliability into the preference flow system, undercutting psephological predictions for the sake of “transparency” and “democracy” or some other idealistic notion.

The lower house was far more traditionally predictable though slightly hindered by the phasing out of tossups, with 133 correct predictions out of 150, or around 89%:

DIVISION PREDICTION RESULT
Adelaide ALP ALP
Aston LIB LIB
Ballarat ALP ALP
Banks LIB LIB
Barker NXT LIB
Barton ALP ALP
Bass LIB ALP
Batman ALP ALP
Bendigo ALP ALP
Bennelong LIB LIB
Berowra LIB LIB
Blair ALP ALP
Blaxland ALP ALP
Bonner LNP LNP
Boothby LIB LIB
Bowman LNP LNP
Braddon LIB ALP
Bradfield LIB LIB
Brand ALP ALP
Brisbane LNP LNP
Bruce ALP ALP
Burt LIB ALP
Calare NAT NAT
Calwell ALP ALP
Canberra ALP ALP
Canning LIB LIB
Capricornia ALP LNP
Casey LIB LIB
Chifley ALP ALP
Chisholm ALP LIB
Cook LIB LIB
Corangamite LIB LIB
Corio ALP ALP
Cowan LIB ALP
Cowper NAT NAT
Cunningham ALP ALP
Curtin LIB LIB
Dawson LNP LNP
Deakin LIB LIB
Denison IND IND
Dickson LNP LNP
Dobell LIB ALP
Dunkley LIB LIB
Durack LIB LIB
Eden-Monaro ALP ALP
Fadden LNP LNP
Fairfax LNP LNP
Farrer LIB LIB
Fenner ALP ALP
Fisher LNP LNP
Flinders LIB LIB
Flynn LNP LNP
Forde LNP LNP
Forrest LIB LIB
Fowler ALP ALP
Franklin ALP ALP
Fremantle ALP ALP
Gellibrand ALP ALP
Gilmore LIB LIB
Gippsland NAT NAT
Goldstein LIB LIB
Gorton ALP ALP
Grayndler ALP ALP
Greenway ALP ALP
Grey NXT LIB
Griffith ALP ALP
Groom LNP LNP
Hasluck LIB LIB
Herbert LNP ALP
Higgins LIB LIB
Hindmarsh ALP ALP
Hinkler LNP LNP
Holt ALP ALP
Hotham ALP ALP
Hughes LIB LIB
Hume LIB LIB
Hunter ALP ALP
Indi IND IND
Isaacs ALP ALP
Jagajaga ALP ALP
Kennedy KAP KAP
Kingsford Smith ALP ALP
Kingston ALP ALP
Kooyong LIB LIB
La Trobe LIB LIB
Lalor ALP ALP
Leichhardt LNP LNP
Lilley ALP ALP
Lindsay LIB ALP
Lingiari ALP ALP
Longman LNP ALP
Lyne NAT NAT
Lyons ALP ALP
Macarthur LIB ALP
Mackellar LIB LIB
Macquarie ALP ALP
Makin ALP ALP
Mallee NAT NAT
Maranoa LNP LNP
Maribyrnong ALP ALP
Mayo NXT NXT
McEwen ALP ALP
McMahon ALP ALP
McMillan LIB LIB
McPherson LNP LNP
Melbourne GRN GRN
Melbourne Ports ALP ALP
Menzies LIB LIB
Mitchell LIB LIB
Moncrieff LNP LNP
Moore LIB LIB
Moreton ALP ALP
Murray LIB NAT
New England IND NAT
Newcastle ALP ALP
North Sydney LIB LIB
O'Connor LIB LIB
Oxley ALP ALP
Page ALP NAT
Parkes NAT NAT
Parramatta ALP ALP
Paterson ALP ALP
Pearce LIB LIB
Perth ALP ALP
Petrie ALP LNP
Port Adelaide ALP ALP
Rankin ALP ALP
Reid LIB LIB
Richmond ALP ALP
Riverina NAT NAT
Robertson LIB LIB
Ryan LNP LNP
Scullin ALP ALP
Shortland ALP ALP
Solomon ALP ALP
Stirling LIB LIB
Sturt LIB LIB
Swan LIB LIB
Sydney ALP ALP
Tangney LIB LIB
Wakefield ALP ALP
Wannon LIB LIB
Warringah LIB LIB
Watson ALP ALP
Wentworth LIB LIB
Werriwa ALP ALP
Whitlam ALP ALP
Wide Bay LNP LNP
Wills ALP ALP
Wright LNP LNP

The senate result...

AUSTRALIAN CAPITAL TERRITORY
Predicted: 1 ALP, 1 LIB
Result: 1 ALP, 1 LIB
Discussion: Exactly as predicted. In the territories, there are only two seats. It's pretty much a given that one major party will win the first, and very rare that the other major party cannot scrape together the rest.

NEW SOUTH WALES
Predicted: 5 ALP, 4 LIB, 1GRN, 1 LDP, 1 PUP
Result: 4 ALP, 3 LIB, 2 NAT, 1 GRN, 1 DLP, 1 ONP
Discussion: 9 right, 3 wrong. One Nation took the predicted Palmer United seat, from a combination of Palmer's personal unpopularity and whatever you personally believe motivates Pauline Hanson supporters: national pride or racism. Labor and Liberal both performed worse than expected, with the Nationals picking up the seats in a state that has always been very strong for them.

NORTHERN TERRITORY
Predicted: 1 ALP, 1 LIB
Result: 1 ALP, 1 LIB
Discussion: Exactly as predicted. In the territories, there are only two seats. It's pretty much a given that one major party will win the first, and very rare that the other major party cannot scrape together the rest.

QUEENSLAND
Predicted: 5 LNP, 4 ALP, 1 GRN, 1 GLT, 1 KAP
Result: 5 LNP, 4 ALP, 2 ONP, 1 GRN
Discussion: While correctly calling the major parties and greens, the minor parties were wrongly predicted. I had considered One Nation a possible contender to take Glen Lazarus's senate seat, their home-state advantages canceling each other out and more or less decided by coin flip. I was genuinely surprised the ONP managed to also dislodge Katter's candidate in their home state too.

SOUTH AUSTRALIA
Predicted: 4 ALP, 4 LIB, 2 NXT, 1 FFP, 1 GRN
Result: 4 LIB, 3 ALP, 3 NXT, 1 FFP, 1 GRN
Discussion: The only correction required would be to switch out a Labor seat for a Xenophon seat. This just goes to show you should never underestimate Nick Xenophon in his home state.

TASMANIA
Predicted: 5 ALP, 4 LIB, 2 GRN, 1 JLN
Result: 5 ALP, 4 LIB, 2 GRN, 1 JLN
Discussion: Called it!

VICTORIA
Predicted: 5 ALP, 5 LIB, 2 GRN
Result: 4 ALP, 4 LIB, 2 GRN, 1 DHJ, 1 NAT
Discussion: Hand one Liberal prediction to their Coalition partners and one Labor to Derryn Hinch. In the first case I should have given more credit to the Nationals in such a traditionally strong state for that party. In the second, I completely underestimated the Derryn Hinch Justice Party who, probably, also got a little luck from the random rollings of the senate's deep ballots.

WESTERN AUSTRALIA
Predicted: 5 LIB, 4 ALP, 1 GRN, 1 NAT, 1 PUP
Result: 5 LIB, 4 ALP, 2 GRN, 1 ONP
Discussion: Looks like I underestimated One Nation across the board. It's the party that just won't die. It's tempting to assume One Nation replaced the waning other minor party from Queensland, the Palmer United Party, which would mean the Greens snatched a seat from the Natss in something of a coup for the left wing of politics. In reality, though, the PUP protest vote probably fell back to the Greens on the whole and the NAT vote stayed on the right with Pauline Hanson. If anything, this is not a left-swing but rather a shift away from the centre to more peripheral or extreme views.

TOTAL ERRORS: 10
ACCURACY: 66/76 (~87%)

I'm honestly quite happy with that result, especially since most of those errors were with regards to the distribution of minor parties which is always a fickle thing. Oh, for the days of the tossup...

All in all those were quite decent predictions in both houses, I feel.

US Presidential

My predictions, however, were not quite so decent here. In short, I backed the wrong horse. However, the predictions were on a state-by-state basis, so lets look at those:

STATE PREDICTION RESULT
ALABAMA TRUMP TRUMP
ALASKA TRUMP TRUMP
ARIZONA TRUMP TRUMP
ARKANSAS TRUMP TRUMP
CALIFORNIA CLINTON CLINTON
COLORADO CLINTON CLINTON
CONNECTICUT CLINTON CLINTON
DELAWARE CLINTON CLINTON
DISTRIC OF COLUMBIA CLINTON CLINTON
FLORIDA CLINTON TRUMP
GEORGIA TRUMP TRUMP
HAWAII CLINTON CLINTON
IDAHO TRUMP TRUMP
ILLINOIS CLINTON CLINTON
INDIANA TRUMP TRUMP
IOWA TRUMP TRUMP
KANSAS TRUMP TRUMP
KENTUCKY TRUMP TRUMP
LOUISIANA TRUMP TRUMP
MAINE CLINTON CLINTON
MARYLAND CLINTON CLINTON
MASSACHUSETTS CLINTON CLINTON
MICHIGAN CLINTON TRUMP
MINNESOTA CLINTON CLINTON
MISSISSIPPI TRUMP TRUMP
MISSOURI TRUMP TRUMP
MONTANA TRUMP TRUMP
NEBRASKA TRUMP TRUMP
NEVADA TRUMP CLINTON
NEW HAMPSHIRE CLINTON CLINTON
NEW JERSEY CLINTON CLINTON
NEW MEXICO CLINTON CLINTON
NEW YORK CLINTON CLINTON
NORTH CAROLINA TRUMP TRUMP
NORTH DAKOTA TRUMP TRUMP
OHIO CLINTON TRUMP
OKLAHOMA TRUMP TRUMP
OREGON CLINTON CLINTON
PENNSYLVANIA CLINTON TRUMP
RHODE ISLAND CLINTON CLINTON
SOUTH CAROLINA TRUMP TRUMP
SOUTH DAKOTA TRUMP TRUMP
TENNESSEE TRUMP TRUMP
TEXAS TRUMP TRUMP
UTAH TRUMP TRUMP
VERMONT CLINTON CLINTON
VIRGINIA CLINTON CLINTON
WASHINGTON CLINTON CLINTON
WEST VIRGINIA TRUMP TRUMP
WISCONSIN CLINTON TRUMP
WYOMING TRUMP TRUMP

Now the good news, psephologically speaking, is I got 44 states right (88%), and 51/57 if you count Maine 1-4 and Nebraska 1-5 as separate races. The bad news is that I got some critical states wrong, such as Florida (29 electoral college seats) and Pennsylvania (20 seats) which would have brought the margin between the candidates within 7 seats of a reversal.

As I said before regarding the Brexit vote, it's very easy to blame bad polling. And there is no doubt that there was a hidden Trump vote for one reason or another. But if the polling is so unreliable then we should either find a better source of data or, at the very least, not rely on it so heavily. Still, 88% is not a terrible result.

The Ghost of Elections Yet to Come


The Infographinomicon will be back in 2017, with a much lighter load of elections to cover; domestically only WA has an election officially scheduled for 2017 (March 11, 2017 to be exact), although both the Queensland and Tasmanian state elections may be held before their 2018 deadlines.

Internationally there are the normal, annual fixtures like the UK local elections and US Gubernatorial cycle. Closer to home, New Zealand will hold a general election some time in 2017. We're also due for another Indian election, and while I didn't cover it last time I do remember the scale of it. Is it odd that I can say “I feel like we just had an Indian presidential election” honestly? Then there are the South Korean and Iranian presidential elections, both of which will be interesting to watch with regards to the political landscape that Donald Trump will be interacting with in the South China Sea and Middle East respectively. Plus, of course, the interesting impeachment revelations of the South Korean president add an interesting kick to the election.

Normally I don't get time to look at these elections, but with a light load next year, who knows?