Thursday, March 20, 2008

What odds a hot 2008?

About 20%, if you just want the executive summary (at least, that is the market "consensus"). If you want the details, read on...

Via Chris Masse at Midas Oracle, I see that Intrade has a betting market on 2008 global mean temperature being in the top 5 hottest years. As Chris observes, the contract is perhaps a bit vague but for the purpose of this post I will assume they will use the global land-ocean index analysis of NASA GISS, meaning the contract pays for a Jan-Dec mean of 0.55 or more.

The current price of the contract is $20 per $100 (bid-offer spread of $15-$25), ie an implied odds of 20%. At first glance, that felt cheap to me, so I had a closer look. 8 years out of the last 10 have satisfied the "top 5" test, which is not too surprising given the overall warming trend. However, temperatures are now in for Jan and Feb 2008, and they are very cold, at +0.12 and +0.26 respectively (note how "very cold" is still well above the 1951-1980 baseline). So it will require a mean anomaly of 0.62 for the remaining 10 months to trigger the payout. That has happened precisely once before, in 2005. Clearly it is not impossible, but it can't be considered likely. We can do a little more analysis to look at the persistence of monthly anomalies. After detrending, the typical e-folding decorrelation time scale of anomalies looks like 4 months or so, suggesting that the relatively cold temperatures will persist for some time to come. That means the latter part of the year will have to really heat up to bring the average up to the top 5 threshold. I'd be more tempted to sell than buy on the contract at the current price, although I've not actually checked the seasonal predictions of modelling centres which could influence my attitude.

Meanwhile, in a parallel universe, some people are making noises about "falsifying global warming" by looking at the last few years of temperature data. I suppose one could perhaps ask what observations would falsify gravity (and for that matter, such observations have been made, although AIUI the interpretation is still disputed). Observations of all LW radiation passing straight through CO2 with no absorption would certainly falsify the theory, but I don't expect to see that any time soon. More seriously, the IPCC explicitly predicted a trend of about 0.2C per decade over the next 30 years, if the trend is substantially different from that then (in the absence of some strong external factor) certainly we'll need to revise some aspects of our calculations. But readers may recall that the Hadley Centre have already explicitly predicted a couple more below-trend years before a continuation of the general warming trend, so I wouldn't get too excited about a few "cold" years, especially when "cold" actually mean "far warmer than most of the historical record, albeit not breaking any global records".



Smart work about global warming. I have also a blog which give information about cause of global warming.

crandles said...

So a single bet of $8 vs $2 makes a 'market "consensus" ' ????

James Annan said...


By definition it's the consensus of that market at this time. Note that even though there's only been one trade, there are offers on the table to buy and sell which bracket that price quite closely, which means that there is not one single trader who disagrees strongly enough to actually take up these offers. I agree that lack of trading and low liquidity is a problem, but I can only take the situation as it is. I also think the price is in the right ballpark, perhaps a bit high (but I haven't looked at any long-term weather/seasonal forecasts). Are you tempted to dabble in this market?

EliRabett said...

Being somewhat lazy, what's the latest on the solar cycle. I remember that the first sun spots showed up a couple of weeks ago. I think, before I bet, I would want a look at that.

James Annan said...


I can out-lazy you any day. But I wouldn't have taken you for a solar theorist :-)


crandles said...

>Are you tempted to dabble in this market?

At the same time as making my last post, I added an order to sell 2 at 23.

Why was I tempted into doing this? Well I would like to see active markets in things like this in order to see what the consensus is (that is just copycatting your thoughts). But why should I think I should be able to benefit from seeing the consensus if I am not prepared to trade myself?

>bracket that price quite closely

11.1% to 23% seem like quite a large range to me.

>there are offers on the table to buy and sell

Yes all of 16 to buy and 21 to sell. wow! OTOH it is early days yet. I hadn't seen the contract listed til I saw your post.

James Annan said...

Someone has pulled their $15 bid since I posted. But yeah, I think a ~5 percentage point range either side of the mid-market price is a pretty tight consensus. Sure, there are only a handful playing - but not one of them is prepared to bid outside that ~10-point range. One could argue that this is a tighter consensus than the situation in a liquid market where equal numbers are trying to push the price strongly higher and lower respectively. IMO the implied odds on some FX claims are far more than 5 points off a reasonable probability, but I can't shift the market by myself. However, no-one is even trying to move this market.

crandles said...

>However, no-one is even trying to move this market.

It seems quite plausible there are only 3 traders that have placed orders - myself and the two who traded the one bet. (This seems plausible because of the regular intervals and .9 .1 endinds of all the orders except for mine.)

No-one out of three people means a lot to you?

EliRabett said...

Semi-seriously, since everything has an effect, combining the absolute bottom of a solar cycle with a La Nina will cool things off, and if you are coming out of both it will heat things up.

crandles said...

I did take a look at the graph towards the end of

1 out of 22 models get to positive el nino values before august.

James Annan said...

No-one out of three people means a lot to you?

Not a lot, but it's not just three - untold thousands have read my post, checked out the price and decided that the price is pretty much right.

Ok, no-one out of 6 then :-)

crandles said...

But maybe Poonam and Eli are in the process of depositing dollars so they can bet..... so no one out of 4... ;-)

More seriously, yes there must be several intrade traders that have noticed the new contracts and decided not to place an order. However I think there is a tendancy to approach a new contract cautiously in the hope that there are some fools out there.

I expect that either nothing much will happen or more people will get involved and the range will narrow. Either way it is too soon to say there is a consensus.

crandles said...

Where are we now that there has been 1 coupon traded at 20, 2 at 23, 1 at 24.9, 1 at 11.1 and 2 at 8.1?

(Apart from a long way from my suggested narrowing range.)

James Annan said...

Sadly my credit cards don't seem to be accepted or else I'd join in.

Brian said...

I'm curious what the price should be if one assumed no warming compared the 20th Century average and a random walked influenced by red noise from recent years.

I can't calculate it but I expect it's lower than 20%, meaning the "vast" Intrade market generally expects temps to increase.

Too bad they don't have a longer term prediction up there, it would make my betting life much easier.

crandles said...

James, was your attempted deposit between $25 and $250?

If you had persevered you might have beaten me to selling 11 coupons at 28.

At time of writing, there is still 12 available at 22.5 if anyone wants to sell at that price.

James Annan said...


I tried a few things and two different credit cards. It warns that many US banks refuse to deal with them and I wouldn't be surprised if Japanese ones are the same. There are increasingly petty money-laundering "anti-terrorism" laws here, plus gambling is illegal anyway.

It's not as if I was going to make my fortune out of it, but good luck with your bets!

crandles said...

What probability now?

I calculate that the last two months would have to average 1.26C above 51-80 base period. Highest monthly anomaly in Nov or Dec is 0.69C and highest monthly anomaly 0.85C. When a record monthly anomaly is beaten I haven't found any recent examples of the anomaly being beaten by more than .21C.

Average anomaly to Oct 0.408C Nov&Dec average .017 warmer than Jan to Oct. Biggest increase of Nov&Dec over Jan to Oct since 1880 is .348C (compared to .852C required). Std deviation of these changes is 0.200C. So more than 4 standard deviations from average required assuming normal distributions.

So the requirement seems greater than anything seen in last 100ish years but it is hard to rule out the probability being as high as a few percent.

Would you be inclined to rule out the probability being as high as 5%? How about 4% or 3% or 2%. (I don't believe a probability of as high as 1% should be ruled out.)

As I write it is possible to sell 10 $10 coupons at 6.1 and 160 coupons at 6 on intrade. That seems a bit high but maybe you would need to be a gambler to risk $1603 for a potential return of $1683.

BTW I think the rules have been clarified by adding 'The data used to expire this contract will be based on "surface air measurements at meteorological stations and ship and satellite measurements of sea surface temperature", as also used in the GISS Surface Temperature Analysis 2007 Summation

crandles said...

Oh well 161 of book order coupons mentioned in last post now withdrawn so little point in asking if you want to bet with me to allow me to take up another say 30 coupons. Probably not worth it for another 9 coupons.

crandles said...

A Book order for 100 coupons at 6 is back up. So I suppose there could be at least a couple of possibilities if you wanted to join the fun and trusted me:

1 Bet with me and I would take up a larger position.

2. Send me some money and I could do a member to member transfer. You would then have $ on intrade and could take up some other intrade betting.

This choice largely depends on if you want you winnings in Yen, GBP or on an intrade US$ account.

It occurred to me that comparing a 10 month period to a 2 month period was somewhat odd and better constraint may be available. I didn't try lots just 1 comparing same length periods.

By comparing Sept+Oct to Nov+Dec. I ended up with 7 Std Deviations above average being required. Required change 75 when previous 128 years had the following changes:

>26 0
>24 1 1
>22 2 1
>20 3 1
>18 4 1
>16 8 4
>14 10 2
>12 14 4
>10 18 4
>8 26 8
>6 30 4
>4 41 11
>2 48 7
>0 54 6
>-2 62 8
>-4 73 11
>-6 81 8
>-8 93 12
>-10 101 8
>-12 108 7
>-14 115 7
>-16 117 2
>-18 122 5
>-20 122 0
>-22 124 2
>-24 126 2
>-26 127 1
>-28 127 0
>-30 128 1


Doesn't seem to look like an excessively fat tail in the upward direction.

James Annan said...

Hi Chris,

Sorry for the late reply. I agree that the bet looks sound although there is always the issue of risk aversion/nonlinear utility. For the sums involved (and given my difficulties in trying to get registered in the past) I really don't think it's worth the bother of getting involved, but thanks anyway.