It’s time for the weekly Intrade roundup for the US Election, so first up we’ll have a look at how the Electoral College vote and Democrat win probabilities have been playing out compared to the Gallup daily tracking poll.
The Obama surge is continuing in both the markets and the polls, and the change in the markets is not only on the headline probabilities and a handful of States, but is general Obama push across nearly all States as we can see if we look at the current State probabilities and their change over the past week:
New Hampshire is the only real piece of good news for the Republicans this week, but that plunge was based on a single trade, so could very easily wash out of the system over the next day or so.
Intrade have produced a spiffy little widget of the State data which I’ll now use as the US election map. A small version is over on the sidebar to the right; a larger version is on the US Election page. It’s updated in real time so feel free to have a play whenever the urge grabs you.
Our simulations this week look like this:
And finally on the data front out Battleground 131 chart today comes in as this:
As always, for greater detail of what all this means and even more data, head over to the US Election page where it’s explained in depth and where a large version of that Intrade map widget will reside permanently.
I wonder whether there is a substantial benefit conferred on the party that has it’s convention first? We know that US polls nearly always have a convention bounce for each party that washes out after a week or so, but for the party that has its convention first (in this case the Democrats), it ends up gaining the very last piece of momentum related to the party conventions simply as a result of the Republican bounce (the party that had its convention last) washing out of the system – but washing out of the system as the candidates head into the final straight.
UPDATE:
Here’s an Intrade Electoral Pendulum. It lists each state according to it’s Democrat win probability (with the probability % given at the end of each bar), as well as the cumulative Electoral College Votes that the Democrats would gain by winning each state in order on the Pendulum.
31 Comments
Possum, with the Intrade Vote Simulation – Frequency View, the longest red line is a 269-all tie, instead of what it looks like, which is a Republican victory. Can you change the colour of that line to avoid confusion?
I might change the colours to avoid confusion – the red and the blue don’t actually represent the Republicans and the Democrats, the entire chart is just the simulation of Democrat ECV probabilities and the blue represents the probabilities above 270 Electoral college votes.
Sorry Possum but I haven’t got a clue why you are saying that the blue and red don’t represent the Democrats and Republicans. The blue portion represents all of the simulations that give the Democrats 270+ votes and the red portion should be 268 or less. Since the election is (effectively) a two horse race, I don’t see any problem.
The simulation is run using only Democrat probabilities. If we did it using only Republican probabilities it would be very very close to a mirror image (because sometimes the probabilities dont add up to exactly 100% – usually as a result of trading on the probabilities of one party being in the process of catching up to the probabilities of the other party… think of it as permanent flux)
So the Red represents rough simulation results of the Republicans and Blue represents exact simulation probabilities for the Democrats for either size of 270 votes.
That frequency histogram – the two longest lines are actually blue ones up around at 364 and 375 ECV’s. Even though they (and for instance the two long red lines) represent high the highest probabilities for any given 4 electoral college vote numbers – each of those 4 long lines only has between a 3 and 4.5% chance of happening. What’s really important isn’t each line individually, but the cumulative results of those lines – which is the second simulation chart.
Just check the chart again and let me know if the big blue lines at 364 and 375 arent showing? I just checked with two browsers and it seems to be working, but would be good to make sure!
Poss that post was such a pleasure to perouse.
Did Nate Silver reply to the detailed comments you left at 538 a few threads back?
Ta Ecky – no he didnt.. well not that I’ve seen anyway.
He’s making the same mistake that most make with Intrade – looking at the wrong markets.
So saying – you watch how the Contender markets start surging to meet the Party markets over the next week or two. Playing silly buggers (or early settlements, or trying to game Clinton contracts etc) can only hold off the weight of money for so long and the weight of money is getting pretty heavy.
Possum,
Hate to say it, but your headline brings to mind the comment about “famous last words”.
The suspense is starting to kill me, and I’m not even American.
News of a surge is welcome, and I hope the Dems keep it mounting right till the election.
Gobama!
Aww Judge Growler – ye of little Intrade faith!
Seriously though, if you look at the types of States that have surged to Obama over the last 2 weeks, it’s included all of the States that usually fall into the punditariats “in play” handbag – with even places like Florida, which today moved above a 50% Dem probability.
It’s not so much the weight of the movement that makes it such a big surge (although the weight is substantial by any yardstick), but where it happened.
The only way you can describe it is in “all the right places” if you’re a Democrat.
pOssum,
You might be aware that we had the festival of the boot down in Melbourne over the weekend.
In a two team race the $1.30 cat was beaten by the $2.90 underhawk. Nothing new about that except all the popular pundits, weight of money, inside knowledge etc. etc. all said a certain result would occur and it did not.
This election is still alive.
Yeah, maybe Greeny – but for the Republicans to win would require something that has never been done before in either the literally hundreds of Intrade election markets that have ever been done or the weight of polls from this far out.
Possum,
Geelong had won 23 of 24 games played this year plus about 8 or 9 from the previous season.
Just because something has happened in the past does not exclude it’s possibility or liklihood.
Just because punters backed Obama early does not yet make them geraniums.
I completely agree. But the evidence is compelling none the less – not only with the punters on Intrade (which have historically out performed the polls when it comes down to the individual states – even this far out from E-Day) but the sheer weight of polling. There are nearly twice as many polls running this cycle than last, and the weight behind Obama is enormous at the moment. Obama really needs to completely crash and burn to lose.
And the public and the punters dont seem to give that possibility much of a chance.
And it’s not hard to see why, especially since Obama’s true level of public support is more likely than not to be underestimated in the polling by a percent or two because of his youth vote and the mobile phone issue.
Greenie – the sun has always risen, there is a small chance that it may not rise tomorrow, in defiance of all precedent, but t is a very very remote chance. An election is not like a football match, where a rather simple objective is attained by actions under the direct influence of the opposing teams. Unless you are in Zimbabwe, (or your brother is governor of the right state and daddy appointed his mates to the supreme court), or something similar, neither side in an election has the ability to directly affect the outcome. The best they can do is kick the ball and see if it dribbles across the line at the right end of the field for them to win.
McCain has about the same chance of becoming President as I have of marrying Miss World.
Greensborough: Nope, at 1.30/2.90 the markets didn’t say a certain result would definitely occur, they said it was an (approximately, neglecting a draw) 70% probability of occuring. With those odds, we’d expect the underdog to triumph about 3 in 10 times. You can’t judge it on a single event – that would be like watching one hand of poker, seeing a royal flush and concluding that the something is wrong because the probabilities say it should almost never happen.
Possum: Simon Jackman has a link to some research on “Cell-phone only respondents” here. Sounds like the effect is well within the usual MOE (although it would of course be a systematic error rather than a random one).
Caf – Pollster.com has looked at it fairly in depth a few times, this being the last:
http://www.pollster.com/blogs/cell_phone_only_households_by.php
It suggests a few percentage points. But that’s at the national level, when you break it down to States and mobile phone use, it looks like the effect will be pretty large in key States – a lot of which happen to be ones that Obama leads only slightly in.
It’s worth a look and following the links there.
I don’t know if I can quite believe that people are analogising an election with a football game. It’s all very scientific.
Caf,
The question is whether the $1.30/$2.90 was a true reflection of the odds. To me the Intrade market is just a giant tote with the price being a reflection of the weight of money. As the Hawthorn/Geelong result showed this is not necessarily the deciding factor.
Similarly, with the US election, Republican strengths seem to have been discounted in a flurry of the crowd flocking to Obama and this is underpinning his price. In punting the crowd favourite loses 2 out of three. Obama’s price is classical mug money. IMHO anyone taking less than $1.80 for Obama is taking under the odds.
That’s true for sports. Not so for politics.
It’s very VERY rare for a favourite in a liquid political betting market (such as the Intrade market for President) to lose. This is particularly so in the US’ case due to the vast volumes of bets that are being placed on this election.
It’s worth remembering that Centrebet have never got a Federal election wrong and have got most Australian state elections right (although the recent WA State election was a notable exception)…
Obama is now strikingly close in the various charts to where he was around 2-5 September (depending on chart), just after his own convention and before McCain had got any real bounce from his. So in effect, McCain has gone backwards by a convention bounce from a position that was unenviable anyway and lost almost half the time he had left before the election in the process.
The Intrade market isn’t a giant tote, and doesn’t act like one. Furthermore, the Hawthorn/Geelong result showed nothing at all – the fundamental difference here is that the odds are saying this result is more likely, not this result is a dead certainty.
Team Obama have a much better ground game than Hawthorn.
Even mugs who attempt to conflate US Presidential Elections with AFL Grand Finals should be capable of appreciating the statistical subtleties at play in these galactically different human endeavours, as Swing Lowe and caf very patiently point out.
Bad mugs of course are another matter entirely.
For those who can’t remember the last time the market got it wrong in and election, it was the last WA election held a few weeks ago.
Ecky,
Please explain these magic mushroom “Statistical subtleties”. I’ll appreciate a good laugh at your contortions.
Greeny – Australian bookie markets are far different to Intrade markets on the US Election. One is thick and liquid, the other is a waste of time. It’s why I don’t use betting data for analysis of Australian politics. Intrade on the other hand has a record that speaks for itself.
GG: You still don’t *get it*. You can’t judge the market as wrong or right on one result, when it was telling you a probability, not giving you a solid gold guarantee. The accuracy of a probability estimate can only be determined over a large number of trials.
caf,
Obviously I’m thick and liquid after yet another long lunch.
However, this current Intrade probability of 62% equates to a market price of $1.60 for Obama winning the election atm. My view is simply that is too low a price and does not accurately reflect Obama’s true price which is about $1.85 IMHO.
So, I remain sceptical about the underlying matters that are driving the lower price. I suspect it is being gamed.
Please note I have previously asked Ecky to provide the basis of the statistical subtleties. Perhaps you can help. I need a good giggle with my dessert.
Your view on the relative probabilities of Obama and McCain is one thing – I’m sure there are many, many traders on Intrade and elsewhere who agree with you, and a decent chunk who think the true value is even further out than that. That’s not my argument – go pick it with someone else.
I am, however, taking issue with you using one result (WA) as evidence that the probability estimates provided by the markets are wrong. Probability estimates cannot be fairly judged on one trial, just like an election result can’t be foretold by polling only one elector. If you go to the casino, play one spin of roulette and happen to come out in front, does that mean that the casino has miscalculated its odds?
Don’t try to drag me into your arguments with others.
caf,
Well who’s the precious one? No more red cordial for yo.
I agree that one result in isolation neither proves nor disproves the of the probability prediction of that event.
What I was reacting to is the unhealthy reverence for the Intrade reliability as a predictor of the true state of play in the Obama situation.
(Apologies if someone has already posted this, I’ve just woke up and haven’t caught up on overnight posts yet)
McCain calls stumps in Michigan!!!
http://www.huffingtonpost.com/2008/10/02/mccain-pulling-out-of-mic_n_131287.html
That’s interesting since the size of the Michigan poll surge has been smack bang in the middle of that group of usually contested states (Ohio, Iowa, Florida, N,Carolina, Colorado etc).
It’s another piece of evidence that suggests the poll surge is more real than ethereal.
Darn right, Poss, *wink* doggone it!
The maverick hockey-momism *wink* of Moose Mama’s launch that surged polls a mere 5 weeks ago *wink* has fast faded in the ice hockey playing northern states of the lower 48. You betcha, *wink*
Little chance of Yup-Yup scoring a political slap-shot now. Pity the Alaskan bears and wolves that will be felled by her bullets of retributative exasperation when Governor Palin exceeds her Nov. 4 GOP Big Picture use-by date.
http://www.first-draft.com/images/2008/04/18/lolcats_feature03.jpg