For the weekly changes in the State probabilities, the simulation results, the daily updates and all the usual bits and pieces, you can just click a link over to the US Election page to see them.
We know from previous Intrade markets on US political contests that probabilities of less than 30% and greater than 70% are as close to dead certainties as we’ll ever see, with no political contestant ever winning an election with a probability of less than 30% the day before Election Day.
So today, we’ll have a closer look at just those states operating in the 30% to 70% probability spectrum and the Electoral College votes involved to find out the Intrade road to the Whitehouse for each of the candidates. Also worth keeping in mind here is that the Intrade results from this day in 2004 were only 2 States out compared to the actual 2004 Presidential election results. Wisconsin at this stage in the last electoral cycle had Kerry on a 41.8% probability of winning that State, even though he took Wisconsin’s 10 Electoral College votes on Election Day by a margin of 50 to 49. The other State which was out of whack at this time in the previous cycle was New Mexico which had Kerry on a 64% probability of victory, yet the Republicans went on to win that State by, again, a margin 50/49.
So at this stage in the 2004 election, the Intrade markets were out by 2 States, or 15 Electoral College votes compared to the final election result.
Food for thought.
If we look at those States that have had their probabilities regularly outside of that 30-70 spectrum, they tally up as 185 Electoral College Votes to McCain and 222 to Obama, leaving 131 votes up for grabs.
We can take those remaining states and break them down into three categories:
McCain Leaning States – those States where the Democrats have between a 30% and 40% probability of victory.
Obama Leaning States – those States where the Democrats have between a 60% and 70% probability of victory.
Contested States – those States where the Democrats have between a 40% and 60% probability of victory.
Just click the charts to expand.
If we look at the race to 270 – if McCain wins all his leaning States as well as every contested State, that would result in a 269 Electoral College Vote tie, as Obama’s ‘in the bag’ States of 222 plus his leaning States likewise add up to 269.
So for McCain to win from here, he needs to defend his own leaners, take every one of the contested States as well as pinch an Obama leaning State. For Obama, he only has to hold his own leaners, and take a single contested or McCain leaning State to claim the Presidency.
The Intrade road to the Whitehouse has quite a few routes for Obama, for McCain it is North Carolina, Florida, Nevada, New Hampshire, Virginia, Ohio and one of either Colorado, Pennsylvania or Michigin.
UPDATE:
We’ll add a new chart over on the Intrade data page that’ll be updated daily along with the Daily Gallup/Intrade Trackers. It’s called Battleground 131 and represents the 131 Electoral College Votes of the 8 States mentioned above that make up the Intrade road to the Whitehouse.
We take the average Intrade probability a Democrat victory of today, yesterday and the day before for each of the 8 states and subtract from this the same 3 day average of one week ago. This gives us a less volatile measure of weekly change for the key States in terms of the probability of the Democrats winning each of these States.
We then plot those weekly probability change values (where a negative score means the State moved toward the Republicans and a positive score means it moved toward the Democrats over the past week) against the current Intrade probabilities of the Democrats winning each of those States we get. (click to expand)
The States are marked along with their respective Electoral College Votes. So if you’re a Republican supporter, you want to see 85 of those Electoral College Votes move below the 50% mark, meaning you’ll want to see at least 85 ECVs worth of States consistently move to the left over the next few weeks. For Democrat supporters you’ll want to see 48 ECV’s worth of States to stay above 50% and consistently move to the right of the chart.
UPDATE 2:
For a bit of fun, we can also use a google time motion gadget to trace that chart through time since June 1st – you’ll need a flash enabled browser to use it.
On the bottom Axis, click on the arrow next to Electoral College Votes and change that to Probability Change over past Week. Make sure the left axis is Intrade Democrat Probability of Victory. Where it says colour, change that to Intrade Democrat Democrat Probability of Victory as well, uncheck the tails box, Under Size choose “Electoral College Votes”. Then click each of the circles to highlight their name and press play. Or alternatively, play with it however you like!
29 Comments
Super analysis as always, Poss. The last 3 or four days have been loking very good for The Kid. Mc Cain’s campaign is pouring money into defending Indiana (Salon War Room) and BHO is getting favourable polls in Florida, Virginia, Colorado, New Mexico and Nevada.
Palin’s pull outside the GOP’s base has tanked and a clear majority trust Obi more than Johnny Bomb-Bomb on “economic management potential”.
A ~60% ROI at centrebet can’t match the value currently on offer at Intrade where around 52 cents will return a buck if The Kid gets the nod Nov. 4.
Have you checked out today’s polls Ecky? Lot’s of blue in them, especially for the McCain Intrade leaners and the Intrade contested States.
Surely we have learnt recently not to assume that financial markets know more than actual people? The people putting their money in the markets simply use the same polling data we have available, making the markets just a poor lagging indicator of the polls, with the added problems of expectation bubbles.
Also, do these numbers take into account the system in Maine and Nebraska? Sure, these may be fairly safe states for Obama and McCain respectively, but unless there is data at the Congressional district level you can’t rule out McCain picking up a single Elector in Maine or Obama picking up one in Nebraska.
Adrian – these aren’t financial markets so they operate a little differently (in the same way that financials operate a little differently to commodities). But in the Polls vs Markets debate as it applies to the US, the markets win hands down on accuracy and predictive power unless your’e a fivethirtyeight.com type that analyses polling data beyond their static nature.
Also worth mentioning here is that for the first 6 months of this year, the polls weren’t a long term leading indicators of the markets, but the opposite was certainly the case. It’s only since the conventions, and on a window of about a week that the polls are starting to lead the markets – which I’ll have a bit to say about in another post next week I think.
Maine and Nebraska have always gone homogeneous in their ECV allocations at elections. Maybe they wont this time – but it’s really neither here nor there for the polls cant really tell us either.
I know you dont like markets Adrian, but they are valuable pieces of data for US Elections by any yardstick – their history speaks for their capability.
The market is essentially a small-scale poll with unknown weighting which asks “who do you think will win?” There is no doubt that it will be roughly accurate, but anyone educated in the field could have predicted at the very start the rough outcome. As both the NT and WA elections showed, people are less accurate at guessing who will win (and hence, markets are less accurate in pricing a win) than large-scale polling data asking people who they themselves will vote for. The markets also bounce around far more than relatively stable polls would justify. Perhaps markets are useful at the start of a campaign where big changes can occur rapidly and state-based poll information is rare, but in the final two months of the campaign this is not the case.
I poorly expressed myself on the Maine/Nebraska issue. Rather than questioning your numbers I was more questioning your conclusions “So for McCain to win from here, he needs to defend his own leaners, take every one of the contested States as well as pinch an Obama leaning State.” Actually McCain could win by defending leaners, taking contested, and stealing a single Elector from Maine. The latest poll for Maine is 50% Obama 46% McCain but it is fairly easy to see how these numbers divided by three could give single Elector to McCain.
Kerry scored 252. Obama is looking secure in Colorado, New Mexico and Iowa, he’ll already be on 273. Take your pick from Nevada, Ohio, Virginia or Florida for the icing on the cake.
In the same way that Johnnie was defending electorates on a 7-10% margin last year, McCain is now pouring his resources into defending states like Indiana and Virginia (which Bush won by 20% & 9% respectively in 2004) haven’t been Dem states for ages, but are now on the knife-edge.
Barring a major blunder, it’s pretty much over for McCain. The GOP vote-rigging systems aren’t going to be a match for BHO.
Adrian – prediction markets don’t work well in Australia because they’re thinner than Kate Moss on a laxative bender (as I’ve said regularly), and bookies have been known to deliberately distort odds to encourage money. Intrade is thick, it isnt like a betting agency or bookie of the type we use here because there isnt an artificial setting of the odds, the whole things works on the price of contracts people are willing to buy and sell from their own initaitve and price setting.
That’s why I pay nearly no attention whatsoever to Australian odds, and didnt last year throughout the election.But for the US Election markets it’s a very different story.The other thing worth mentioning is that the political organisations on the ground in the US and the participants in the political system are much much larger in number than they are here, and as a result the information about the political state of play in any given state is known by a much larger group of people – and we see that come out in the State markets regularly, with the simulations I do using those state markets often being leading indicators of the broader national headline market, or at the very least pointing to an error correction direction of the national market.
In the 2004 presidential election, in the 2004 congressional and senate races, in the 2006 mid terms, Intrade broke earlier, and more accurately than did the polls. That’s the reality of the data. When the polls were still showing lot’s of uncertainty or sampling error giving wide margins of error because of poor poll size, Intrade was much more stable in the last few weeks of each of those campaigns.
Why? Good question. But we cant deny that it actually happened – and writing Intrade off as being a mere follower of the polls isn’t actually consistent with the state of the polls during those contests…. nor this one for that matter.
This election it’s a bit better with polling info because of people like Nate Silver and the work he does, but there are still a few disconnects between current polling and Intrade probabilities. Florida being a case in point.
What’s the bet Intrade will be right again? (boom tish)
Poss, you may have covered this already, sorry if you have, but how are yu dividing up the ECV votes, is it winner takes all for all states? I seem to remember from second year politics with Malcolm Mackerras, that not all states divide their ECV allocation in an first past the post manner, but some are proportionally allocated, possbly to some mad formula, is that the case, its been a long ten years and my brain hurts.
Much rather be Obama than McCain – it’s not exactly a lay down Misère, but McCain would have to pull of a blinder to win.. and with the current economic Tsunami – McCain would be hard pressed to do a Canute and hold back the tide.
PC: “Have you checked out today’s polls Ecky?”
Possum, two o’clock e.s.t. I posted. You are havink zee leedle joke, non? Often the first injection is administered at dawn. The family feels it’s best for all hosehold members that way:)
Do, however, like this poll from ABC/WaPo who are hardly card-carrying fans of The Kid. Senator Obama has a 9 point spread nationally in this poll.
Not good news for GOPpers at all.
BHO 52, JMc 43.
http://demfromct.dailykos.com/
saul, Maine and Nebraska are the two states that can split their electoral college votes, but in practice they’ve never actually done it. The ultimate Presidential winner of Nebraska and Maine has won all districts in each state since those particular electoral college vote allocation systems were introduced in 1972 for Maine and 1996 for Nebraska.
I allocate in a few different ways over in the simulations, but for this post it was done on a basic winner takes all approach.
DR – same. Much rather be Obama considering those numbers and considering the time to E-Day.
Ecky – I can understand the family’s concern, being a fellow poll addict. I’m getting paranoid about one of those Intervention things that are all the rage with addictions!
“Noooo, Nooooo – just after the next poll I promise.There’s a really important ARG for…er.. Hawaii coming out that I can’t miss!”
That 52/43 was a killer. Wonder if it will dominate the news cycle tomorrow morning US time? Nasty way to start the day if you’re on the Straight Talk Express I’d imagine.
Possum, if you haven’t already, check out Nate’s latest post at Fivethirtyeight.com. Looks like there may be someone playing silly buggers with Intrade at the moment.
http://www.fivethirtyeight.com/2008/09/intrade-betting-is-suspcious.html
Poss
Al got to it before me. 538 has called Intrade out. I’ve been thinking about the possibility of betting markets being manipulated for political gain. The “expectation to win” factor is hugely important in elections. It would be worth McCain’s campaign $1M to keep propping him up in a high profile market. Money well spent. It’s like advertising.
Yeah, well spotted, Al. I noted the arbitrage potential in my initial comment on this thread but hadn’t jerried to the possibility of Intrade “funny business”. That’s why the gods of psephology and statistical soprts analysis have sent Mr. Silver to guide us.
In hindsight though, their market was behaving oddly.
With regard to Teh Furry One’s projections with the recent Intrade numbers, the “true” Intrade figures make The Kid an even better special to bring home the bacon Nov. 5.
As propaganda, the “true figures” would be devastating for GOP esprit de corps if published on RCP homepage and elsewhere. Even rank and file GOPpers would twig that The Kid is close to home and hosed.
I’ve just answered Nate over there – I have a sneaking suspicion I might have been the cause of that post from a comment I left in the previous thread.
I agree with your analysis Poss, but McCain is not quite as constrained as you point out when you say:
“The Intrade road to the Whitehouse has quite a few routes for Obama, for McCain it is North Carolina, Florida, Nevada, New Hampshire, Virginia, Ohio and one of either Colorado, Pennsylvania or Michigin.”
If McCain wins one of CO (9 ECV), PA(21), or MI (17) as you say, then that gives him the lattitude to lose some of his leaners like NH(4), NV(5) and as long as he wins PA or MI, he could even lose VA(13). So he doesn’t really need the slam as you suggest, and I imagine that McCain will throw huge amounts of ad spend at PA and MI precisely so he can have that flexibility.
Don’t think it will work, though. At least, I hope it doesn’t
Well JP – if you want to go and get all complicated! Sheesh
If you can be certain that the < 30 longshots will not come in, you can get a guaranteed 30 – 40 % return in a month and a half. Not bad, to say the least. I was going to punt some savings on MQG and CBA – maybe I’ll spread it around the longshot states on Intrade instead!
I’d really love to see an “Intrade probability of win the day before the election” against “Actual proportion won” plot. My hunch is that you could fit a fairly simple curve to it, and then use *that* curve as an actual evidence-based way of adjusting the Intrade probailities of markets in the future.
Battleground 131 motion time gadgets added!
Hope they work.
Caf – I’m actually hoarding as much Intrade data about this election as I can get my hands on, including all the different simulation results with the aim to create a few Intrade blueprints for dealing with uncertainty in the future.
Poss, have you done any analysis as to how accurate the Intrade-type markets become the closer the election date is? I have no doubt that come election day the pundits putting their money where their mouth is the most accurate predictor we have seen, but how far out does that become the case?
Re: Update 2
Possum you are a genius.
I also loved your extended post on 538.com referenced earlier. I love how you are able to explain your methods and attitudes towards analysis so well. Although i feel i need to read everything twice to understand it, it is well worth it.
Possum,
I have been a lurker on this (and many others) for some time.
I enjoy your posts, though I don’t pretend to understand them well.
Although you seem not to like http://www.fivethirtyeight.com is there anything in the article at the following link?
http://www.fivethirtyeight.com/2008/09/intrade-betting-is-suspcious.html
He seems to be saying that the Intrade betting is being manipulated.
PC, Battleground 131 motion time gadgets are geeksville De-Luxe. Looking forward to another poller-coaster ride, this time maybe through the Grand Canyon when the tallyin’s all done and dusted in November.
What effect will Dubya’s re-emergence from the bunker doing a Hanrahan have on McCain’s chances? I suspect it has just about sunk any chance McCain has, not helped by McCain now going AWOL.
Dario,
Intrade has been more accurate than poll results (often perfectly accurate such as 2004 Pres election) immediately before election day for every set of US Elections they’ve had markets on.
At this time in 2004, they were only 2 States out from the eventual result. A few weeks before E-Day that dropped down to being 1 State out (but remember, for a few hours on one day a State might move from one column into another before the weight of market money moves it back into it’s regular column). So they’re pretty much good to go from here on in.
Thewetmale,
Glad you enjoy. If you ever have any questions, please feel free to ask. That’s what this place is about – trying to make the complicated not so complicated.
Steve,
Don’t get me wrong, I think Nate Silver of 538 is brilliant and he’s certainly my favourite US polling analysis site. He was right in that some Intrade markets are dodgy – he was wrong to suggest that Intrade is dodgy as a result, because he made the mistake of using the wrong Intrade market for analysis.
At Intrade, there’s actually four markets that you can bet on over “who will win the US election”.
The first and most popular market is the Contender markets – the Individual as Winner markets where you can can not only bet on McCain and Obama, but also Clinton, Romney, Gore etc. 18 candidates can be bet on all up.
The second is on which Political Party will win the Presidency.
The third market is the collection of State markets – which party will win which State
The fourth market is the “pick the number of electoral college votes for each party” market.
That first market – the contender market – ceased to be functional and relevant after the nominations for the Democrats and Republicans were sown up. Now, it’s mostly filled with newbie punters and people trying to recoup their losses on earlier contracts. For instance, there was a lot of “Clinton as winner” money floating around in this market and there still is. After election day, all the people that backed Clinton will have to pay up. But Clinton can still be traded, so people seem to have packaged up a series of bets and betting volumes that they initiate at the same time (not too dissimilar to what an Intrade derivative package might look like) with the hope of manipulating the market price for a few hours and reducing the size of the losses they are holding on Clinton. Not by much, but by a few dollars here and a few dollars there every time they game the system.
That’s what Nate picked up on – but that market hasn’t been a serious market for months and isn’t worth paying serious attention to. I’m surprised he used that as a broad representation of Intrade because no one with any serious knowledge of prediction markets or statistical nous ever uses that market for anything!
If the Political Party as Presidential Winner market was a 4000 sample national poll run by a reputable pollster, the Contender markets would be the equivalent of a tabloid newspaper running an online poll on the front of their website.
The Political Party as Winner market is the headline market where the real action is (although still slightly skewed by the Contender market results washing into it a bit).We don’t see anywhere near the blatant gaming attempts in this market that we see in the Contender markets – yet even when they do pop up, their results get washed out of the system usually within the hour.
It’s also worth mentioning here that to manipulate Intrade to any level that makes a real, sustained difference in price and implied probabilities would cost vast sums of money. Vast sums of money. But the more it was attempted to be obviously manipulated, the more people would jump in to invest in that “free money” that would be floating around.
To get the most out of Intrade info – dont look at the Contender markets because they’re legacy markets. Look at the Political Party and State markets instead.
It’s not necessarily “gaming” that’s happening in those Contender markets. Another reason is people wanting to pay a few percent premium in order to close out their positions now, rather than waiting until the election. For example, people that successfully went short on Guiliani still have to wait until the election is over for their backing funds to be unfrozen – unless they buy Guiliani now (very cheaply) to close out their position immediately. That frees up funds for other trades. (The people on the other side of the trades are effectively investing funds at ~ 20% pa for 40 days, so they’re happy too).
“those States where the Democrat’s have between a 40% and 60% probability of victory.”
Possum had better hope Apostrophe Man doesn’t see this atrocity.
Apostrophe Man seems to have realised the futile exercise it would be to try and improve my use of curly punctuation marks
But just for you Dr Carr, I’ve saved your eyeballs from any further assault.