Predicting the 2015 Canadian election

The Canadian general election will be held on 19 October. The most basic prediction method uses the full district (“riding”) vote information from the last election (in 2011), the current poll estimate for national level support for each party, and a model of changes in district votes. There are two main models used in predicting elections under First Past the Post (single-winner plurality in districts), namely Uniform (additive) Swing and Proportional (multiplicative) Swing.

Based on the aggregate poll at signal.thestar.com, these two models predict the following point estimates for the seat distributions (after scaling up to account for the increase in parliament size since 2011):

Multiplicative: CON 133 NDP 71 LIB 125 BQ 7 GRE 1
Additive: CON 145 NDP 85 LIB 101 BQ 6 GRE 1

NDP have lost a lot of support in recent weeks, but it still looks as though no party will have an absolute majority and CON will be the largest party.

UPDATE 19 October (NZ time): using the latest poll estimate the models now give:

Multiplicative: CON 131 NDP 72 LIB 128 BQ 3 GRE 1
Additive: CON 137 NDP 86 LIB 109 BQ 5 GRE 1

ThreeHundredEight.com predict: CON 120, NDP 71, LIB 141, BQ 5, GRE 1
Toronto Star predict: CONS 124, NDP 71, LIB 140, BQ 2, GRE 1

Let’s see the results tomorrow.

PolNet 2015 and APSA 2015

In June and September I attended two contrasting political science conferences on the West Coast of the USA. PolNet 2015 was small (of the order of 100 attendees) and focused on political networks. APSA 2015 was huge (about 7000 attendees), covering all areas of the field, with many business meetings and other professional activities in addition to research talks. Both were very well organized and had unusually good food available. These trips were very tiring physically, but very stimulating intellectually.

Personal highlights:

PolNet:

  • according to Skyler Cranmer, statistical inference on survey data usually makes an assumption of independence, which makes no sense for networks, and he found 700 published papers making this methodological error
  • David Lazer (and many collaborators) Moneybomb video
  • Kathleen Carley: 25-50% of tweeters are not human – biased data for social scientists!
  • really useful introductory tutorials by Skyler Cranmer, Lorien Jasny, Katherine Ognyanova, Scott Pauls

APSA:

  • plenary talk by Robert Reich – in addition to the main substance, the anecdotes were very entertaining, especially his interview by a journalist askingabout a date he went on with Hillary Clinton about 50 years ago
  • British Election Forecasting session – interesting to see how so many different predictions were all so wrong, and concerns about “forecaster herding”
  • political networks session where several people agreed that it is a fad, but that is OK because it gives good results
  • APSA business meeting into which I was press-ganged in order to make a quorum. It seems that elections use Australian House of Representatives rules. The quorum was lost partway through, and a complaint made … The political scientists can sure talk!
  • talk by Carey, Masoud and Reynolds showing how close Tunisia came to failing like Libya and Egypt (small differences in the voting rules used)
  • Alex Montgomery’s favourite international organization is the  African GroundNut Council
  • many other interesting topics: why is there no right-wing party in many countries? ; effect of preferential voting on representation of minorities; poll herding hypothesis – low quality pollsters adjust results to match high quality ones; how can polarization be measured?

 

Barriers to open access – reading

Here is a list of references that may be useful when considering barriers to open access.

Measures of partial balance in signed networks

Networks in which the edges can be positive or negative occur in many applications (for example, in international relations, states may be allied or official enemies). A widely-used theory of “balance” predicts that networks should become more balanced over time. However there are no clearly agreed measures of partial balance. Samin Aref and I made a start on improving this situation.

Abstract: Is the enemy of an enemy necessarily a friend, or a friend of a friend a
friend? If not, to what extent does this tend to hold? Such questions were
formulated in terms of signed (social) networks and necessary and sufficient
conditions for a network to be “balanced” were obtained around 1960. Since then
the idea that signed networks tend over time to become more balanced has been
widely used in several application areas, such as international relations.
However investigation of this hypothesis has been complicated by the lack of a
standard measure of partial balance, since complete balance is almost never
achieved in practice.
We formalise the concept of a measure of partial balance, compare several
known measures on real-world and synthetic datasets, as well as investigating
their axiomatic properties. We use both well-known datasets from the sociology
literature, such as Read’s New Guinean tribes, and much more recent ones
involving senate bill co-sponsorship. The synthetic data involves both
ErdH{o}s-R’enyi and Barab’asi-Albert graphs.
We find that under all our measures, real-world networks are more balanced
than random networks. We also show that some measures behave better than others
in terms of axioms, computational tractability and ability to differentiate
between graphs. We make some recommendations for measures to be used in future
work.

Link to preprint: http://arxiv.org/abs/1509.04037

Distance-based voting rules

After a long gestation period in which I seemed to be publishing nothing, a few projects have reached maturity. With Benjamin Hadjibeyli, I have a preprint studying so-called distance rationalizable voting rules, which we recently submitted. These are voting rules in which we specify some notion of consensus winner, set up a distance measure on elections, and then choose the winner(s) based on minimizing the distance to a consensus election. This framework has been used by other authors over the years, particularly by Edith Elkind, Piotr Faliszewksi and Arkadii Slinko in a series of papers.

We deal with a lot of foundational issues and give some insight into the way Minkowski geometry relates to decisiveness of such rules. This includes both positive and negative results.

This work was inspired by discussions with Elchanan Mossel and Miklos Racz in Berkeley in 2013, about whether the Dodgson rule is a so-called hyperplane rule. That question is still not answered, but now the underbrush has been cleared, it makes sense to go back to it.

Reforming the profession

I have been following several issues in the world of academia and science over the last few years, all of which require urgent and serious attention, in my opinion. Progress has been less than I hoped, but it makes sense that inertia is large in such a system.

I collect here a list of important and interesting links, and hope they are useful for others.

Visionary stuff

Evaluating researchers

Evaluating research

Access to research

Researcher ethics

Inclusive culture

People to follow

Older writings

Experimental research

When I was a PhD student, stretching my horizons meant thinking about commutative ring theory, instead of general rings. Over my career I have gradually stretched further, taking in mathematical parts of computer science and social choice theory. However in recent years the stretching has become much larger. In addition to supervising PhD students in network science, my first (joint) work on experimental social science has been uploaded to the world.

In this work, ultimately inspired by a logical model developed by our colleague Patrick Girard and coauthors to describe belief changes in social networks, Valery Pavlov and I have conducted a laboratory experiment with human participants, designed to measure influence and social learning of factual information. A novelty was the way we allowed and incentivized participants to truthfully report “I don’t know” – this seemingly small change has large effects on the dynamics.

Almost everything about this was new to me, but I now feel confident about taking this work much further. Threshold-type diffusion models, as opposed to the infection-type models so common in network science, seem to be much more relevant to this kind of situation. Our work suggests a different model from the usual threshold model.

Who knows what the next decade will bring? Perhaps an art installation or a musical performance? There are still faculties of the university I haven’t been much involved with.

Canadam 2015

Last week I made a flying visit to Canada for this annual meeting (it’s hard to imagine any other type given the distance, but it was particularly short – less than 72 hours in the country). I was invited by Steve Melczer, an impressive PhD student who seems to have read my coauthored book better than anyone alive, and with whom I hope to collaborate this year. Despite the arduous travel, it was a worthwhile experience and should lead to some nice joint publications.

I noticed that many of the buildings at the beautiful campus of the University of Saskatchewan are connected underground, giving a clue to the winter weather. Even though it was early June, the weather was not particularly warm. On my return via Calgary Airport, I noticed that it was 12 degrees and raining at 4pm (in summer, or close enough). The next morning in Auckland at 4am, it was 13 degrees and raining (in winter).

Rare events

Yesterday was an unusually busy day, but different from most unusually busy days of the last 15 years, because it was mostly fun. I took my younger son to the opening game of the FIFA Under-20 World Cup (NZ vs Ukraine). Despite obvious inadequacies in the logistical planning at the venue, it was a great occasion. Soon after making it home by bus, I went with my wife to NZ Opera’s production of La Cenerentola. Having not been together to such an event for 15 years, we enjoyed it thoroughly.