Something we don’t pay much attention to around this place are the various political betting markets – so from now through until the election, we’ll have Betting Market Friday as a regular post. Essentially, what we’ll do is look at the federal election prices from five betting agencies – Centrebet, IASbet, Sportingbet, Betfair and Sportsbet – and do all sorts of nerdy things with them over time.
As we approach the election, we’ll be able to answer questions on whether poll results lead the betting markets around by the nose, which pollster has the biggest impact on the betting markets, and as individual seat markets open up, we’ll also be able to do some monte carlo simulations on them and see how the predicted election results differ from the latest polling snapshot election simulation.
We’ll also add the betting market implied probabilities for each party into the sidebar as a little graphic – because, well, that sidebar isn’t quite filled with enough information yet 
To start the ball rolling, we have to calculate ourselves some implied probabilities from the prices that the betting agencies are offering. This is a two stage process – firstly we have to get ourselves some nominal probabilities. Secondly, we have to adjust those nominal probabilities for the overround – the effective commission that the betting agency makes from the book.
To calculate the overround and the nominal, unadjusted probabilities – we just simply take the inverse of the prices to get the latter, and sum them together to calculate the former. Using an example, the latest Centrebet prices have the ALP on $1.20 to win the next election, while the Coalition is sitting on $4.25. The calculations then look like this:
These prices give nominal probabilities of 83.3% likelihood of victory for the ALP compared to the Coalition’s 23.5%. However, you may notice that the probabilities add up to 106.86%? That excess probability is how the betting agencies make their money.It represents the cream they take from the market.
If we assume that the betting agencies are making money equally from both sides of the betting equation (in that the ALP probabilities are inflated in the same approximate proportion as the Coalition probabilities), we can adjust for this overround to get at the true implied probabilities associated with the prices by simply dividing the nominal probability of each price by the overround.
Currently, the Centrebet prices imply that there is a 77.98% probability of the ALP winning this coming election compared to the Coalition’s 22.02%.
If we do that for all five betting agencies and calculate an average based on the raw data from those agencies, this is what we end up with.
Our five agency average is what we will primarily use this year to track the markets.
But wait, there’s more! Hold onto your handbags folks, we can graph it!
OK, so it’s not particularly exciting yet – but over time I’m sure that will change.
As we get more data, we’ll also look over the literature – particularly the Australian literature – on betting markets in politics. One of the good things here is that some of our favourite economists pretty much dominate the Australian academic literature with the likes of Andrew Leigh, Justin Wolfers and Simon Jackman being particularly prominent.
Disclaimer: Gamble at own risk. I may be many things, but a fortune teller is not one of them.
UPDATE:
I missed Sportsbet, so that’s now been added into the mix
42 Comments
Thanks for this. I look forward to the graph getting some more points on it.
Any chance of a retrospective on betting agency odds against polling and actual election results? Or a link to same?
Unfortunately Aubrey, I don’t have a very good set of betting market data from 2007.
The biggest mistake I made was not keeping a good dataset on that last campaign.
That’s a spiffy graph you got there Possum!
The good thing about this is that commentators always wheel out the “betting markets are a truer indicator than the polls” line at least 6.75 time each in an election year.
(I can’t vouch for that number – I just took the polynomial trend)
Individual seats could give examples of variance around individual issues,
Many here made a bit of pocket money betting against Howard, and I remember those rumours about someone stumbling out of Lib head quarters and slapping $10,000 on Howard, just to keep the odds somewhat promising.
For a profit we need to see where we have that type of market warping, then I presume we all trot off and bet against it?
Barking, one of the things we can do when those individual betting market seats open is compare their implied probability of victory against the implied probability in the quarterly election simulations using polls.
If one were so inclined, one could bet on a series of seats where the probability difference between the two metrics blows out.
I think Andrew Leigh’s website highlighted betting “opportunities” at the last election where there was a mismatch between seat polls and the betting. At the seat level, public polling is often done with small samples and likely to be volatile. The l ocal betting market could also be vulnerable to insider trading and market manipulation. Clearly, the party workers have a marked advantage in their local knowledge and even though the betting companies try to defend against this advantage, it’s pretty hard to stop rumours spreading!
Howard tried to bribe me in the 2007 budget with $500, I put that $500 with Superbet @ 3.25 for him to lose his seat, I am now the owner of a plasma telly which every time I look at it gives me great satisfaction.
Whoops, correction Sportingbet.
Alas! SA Gambling Authority prevents me betting on the election results! WTF for I dunno!
I too made a mozza on Maxine
Poss,
I think a disclaimer is appropriate. Something like: I take no responsibility for any losses on “investments” based on my analyses.
Feel free to take any credit, though.
Will do vp.
Possum 5.
Thats exactly right.
Again no evidence here but I remember looking through the individual seats at the last fed election and there were cases where a local candidate had better numbers than the actual position. Clearly one of the influences is the ego and candidate fever of individuals.
A local example I saw was nearby seat, it was an alp seat, 51%-52% ISH, the local coalition candidate was absolutely convinced that he was going to romp it in, quite bizarre, his campaign team who he called ‘political geniuses’ were the same, they were all cashed up upper class ‘club’ members, they got onto his odds that were about $5 a win and all put a truckload on him. He had a 5% swing against, and I still get a chuckle.
One of the advantages of the seat analysis would also be a spread of risk, say 5 seats with $1.20 would allow for a break even if one candidate fell for the old, “photo with goat” trick between the punt and the vote.
Last election there was also the $100 introductory offer from some of the betting agencies, if anything similar pops up we should let people know.
I know I’ll seem totally one-eyed saying this but Labor must be the best odds-on bet since Gunsynd was running around Randwick.
The only sensible betting would be about the extent of the Labor victory – ie odds-on for a landslide.
If others do think the Coalition has a reasonable chance then I’d like to see their reasoning.
Okay, this is a really good addition.
We all know that there are essentially three problems with you’re excellent monte carlo analysis;
1. You have to use state level uniform swings – meaning that the prediction of individual seats is limited (though, perhaps evened out over all).
2. You don’t have the data to predict the senate.
3. The monte carlo analysis is ‘as if the election were held that day’.
Fair to say, that I’ve questioned you regularly on 1 and 2 – and you seem to be doing the best you can with the data you can get.
Including betting markets as an indicator, if they are in fact accurate – and it seems they are better than random – is looking foward to an election. This gives as an indication of what people at least are feeling about the potential uncertainty. Therefore, to some degree, this will address 3.
If anyone is interested in the literature I read this sometime ago and it was really interesting.
http://econrsss.anu.edu.au/pdf/DP502.pdf
Maybe you could put this all in a combined model like they do?
The number of people I know who bet on Maxine at the last election makes me wonder how in the heck she wasn’t the favourite!
Grog,
Not only that, does anyone know anyone who backed Howard?
Peter Costello?
This is going to be loads of fun. Thanks Poss.
On the topic of ‘betting markets being more accurate’ is there any way of telling how many bets have been laid? I’m out of my depth here, but my understanding is that the betting market responds to how much money is thrown on one side of the other? So, if we knew how many data points were involved there could be some sort of MoE?
Possum
I’ve been tracking Centrebet since last August; see here
Quite a deal of the data collection & analysis I did re the betting markets in 2007 is still up on my site: e.g., this graph
Thanks Simon.
There’s an update as well folks – I missed Sportsbet, so I’ve added that into the mix.
Possum – Sportsbet’s $1.20 and $4.50 equal different probabilities than IASbet’s same odds. I presume this is a mistake? :p
Aww man, don’t tell me that! (scurries off back to the spreadsheet)
Apols for the brainfart – all fixed.
Thanks for picking it up Max.
Excellent addition. I think Tas Labor and the SA Libs still offer some decent value.
This week’s Morgan is out:
http://www.roymorgan.com/news/polls/2010/4466/
The new Pollytrend and Poll Average charts have now been updated on the sidebar.
Betfair and the others usually operate at about the percentages you have here but the odds should be better to take account of their use of your money for a year or so. You can put $100 on Liberals now for a return of $120 (5 to 1 on) but at the time of the election, in say a year, your $100 on deposit is really about $105, which for a return of $120 is really odds of 7 to 1 on. I bet on elections at times (yes Grog, I also backed Maxine) but I usually leave them alone until at least 6 months out.
Thanks to all the gentlepeople who responded about fiscal / monetary. It is impossible to find a definition of fiscal other tha that it pertains to money and finance (duh). It seems monetary relates to the control of the amount of money circulating within the economy (as controlled by the RBA in this sea-girt country) and fiscal relates to any other measures (some of which can be controlled by the government) that can affect the economy.
Sorry Poss,
That was meant for William’s.
Yep, it was like free money Grog, and seeing Howard “Bruced” was the icing on the cake.
Nice graphs Poss.
EVAN BEAVER: The amount in the winning pool at the races is always announced over the PA. Another way is to stand in the mounting yard enclosure area when the PA announces something like “The pay out for the last triffecta for horse 4, 15 and 6 is $345,000 for one unit of investment. The dead silence following the announcement will tell you how many people didn’t have the winning ticket.
The odds quoted in the lead up to the election will tell you how many people are voting. 4-5 on means practically everyone is voting for more than one party a Party. 100-1 against no one is showing any interest. 300-1 the candidate just died. 9-10 on just forget it.
I would think that the North Shore punters for Howard would have put on far more in $ terms than the average punter such as ourselves on Maxine.
The secret was to get on early before the money started coming in for Maxine. I got $4.60 but stupidly only put on $30 and the free $100 on Labor. I think it only paid $1.05.
I hope you’ve thought this through Possum
I love the idea personally but, well, you know how the blogosphere can get
Another comment, one of the annoying things that occured with the spoils of the Howard $’s was that whilst the betting agency were happy to deduct the moneys straight from the old credit card account, I had to send a letter confirming who I was and stating the bank account I wanted the money depositied into, As I would never bet on anything else I found this pretty blatant delaying tactics, I presume not many punters take their winnings out,
Do other people have similar stories, or are some firms better than others.
Dear Poss,
Will there be enough room in your crowded life to do a similar exercise on the coming election in Victoria? It could end up being interesting.
BTW: My little comments re betting ie at 300-1 against the candidate just died, was just me having some fun.
Venise, I’ll certainly cover the Vic election. In about June/July I’ll start to focus on it a little more closely.
Poss. Good, good, good!
A spokeswoman for the Super Funds national body has today denied that the Mining Super Profit Tax has effected the funds,,,she blames the Euro slump..rightly I think !
It shows the way the Mining Barons will lie and deceive to win the argument
The Richest people in Australia don’t like the Tax…surprise surprise !!
Gina Reinhart is the daughter of that old monster Lang Hancock who before his death married a philipino “lady” Rose Porteous, who got lots of money from his will. I bet Gina didn’t like that either !
Just another relevant post for me. I admit that I don’t really know enough about betting market Friday but this post explains some questions that i had in my mind for almost a long time.
Keep it up.