Wednesday, March 18, 2015

A Rashomon update on climate betting

So 2015 is coming in hot, .75C in January and .79C above base period in February, and nothing indicating that March is cooling. The rest of the year can be slightly colder than 2014 and still end up a record, the first time since 1998 that we had record years back-to-back.

While short term data on climate isn't especially meaningful, I have a particular interest in this - in addition to the minor issue of human and environmental welfare, I've got $9,000 riding on it. David Evans and I made a series of bets in 2007, and 2015 is the first year that starts counting against the baseline. David has the details here, but to sum up it's comparing 5-year averages, and we're betting both on temps warming either somewhat less than IPCC projected for the next few decades or much less (David acknowledges some warming is possible). I need temps to increase .13C/decade over the 2005-2009 baseline to win one bet and not lose/void the other one, and .18C/decade to win both.

The Rashomon aspect comes from whether the bet looks like good news or bad news depending on your focus. I'm winning the first two months of the five year period from 2015 to 2019, which is good for me but not all that definitive. Prior to 2015, comparing years that didn't count, I was losing the bets, and prior to 2014 I was losing them badly.

One way to view it:  if you focus just on the stats and forget all your priors about the science, and if you ignore the stats for the many years preceding the baseline period, then I think you'd rather be in David's shoes than mine, despite my little head start. If 2015 stays warm at the end of this year then you might feel differently, even retaining this constrained viewpoint.

I think it's a reasonable perspective, but too constrained. If you look at a longer period and consider reasonable scientific priors about what you expect to happen, then I think I still have good bets. There's also what I said in 2007 that I "at worst lose one bet, win most of them, and void the rest." I also said I had the best chance of losing the early 10-year bets as opposed to the longer 15 and 20 year bets. Not so different from my expectation.


The other interesting bet to look at is James Annan's, where he's over halfway through the determining period for his bet. He's coy about it but he's trouncing his betting partners who thought temps would actually decline compared to 1998-2003.

A perspective I think would be interesting to hear from is the two men he's betting with. While their economic interest would be to wait and hope for a miraculous change, they have other reputational interests. Rather than continuing to be wrong this year and for the next two years and then finally paying up, I'd end the period of being wrong now if I were them, write James a check, and start being right.

20 comments:

  1. Science priors trump everything.

    It is a non-linear feedback system. Statistics from previous periods are not informative.

    Your risk was not how much heat would accumulate, but how it would partition, into the various phases.

    Given the current levels of all greenhouse gases and albedo effects, there is enough heating that the partitioning issue is likely now moot.

    Do I feel sorry for the poor rubes you are fleecing? No, folks who make bets should know their science. Science always works.

    ReplyDelete
  2. Still no takers for

    ' In which decade are you prepared to bet " on 1/10th of any year 2100 delta T in the IPCC range first being recorded globally ."

    E. G. 'I bet we will see .5 degrees from 2040-50' or 1.1 degrees from 2090-2100. ...

    Exponentiality must start somewhere but no thus far nobody wants to bet on when.


    ReplyDelete
  3. 2060-70 seems like the earliest one could confidently bet in favor of that, Russell, making it an impractical bet for most. Even 2050-60, which might approximate a bet fair for both sides, is a bit too far off to be practical.

    Actually I think carbon feedbacks would make 2050-60 a sucker's bet for you, but even that's too far off.

    But perhaps I misunderstand your suggestion.

    ReplyDelete
  4. Still no takers for...

    May I ask where and when you've previously tried to recruit takers for such a bet? And could you point to one such place where you explained the proposed terms in more detail? The fortyish words above don't really suffice to establish the details of what temperature record or records you'd be willing to bet on nor other details. If you're actually serious about finding a taker, maybe you ought to post a full description of the terms you're offering somewhere easy to find. As it is, googling "Russell Seitz climate bet" and "Russell Seitz temperature bet" doesn't seem to return anything helpful.

    ReplyDelete
  5. The Bunny's Munny is (probably) in the Hutch

    Take a gander at this graph, which shows GISS temperature with a simple 1-variable model:
    https://farm9.staticflickr.com/8642/16244686434_fbd6b1dd1b_o.jpg

    The 1-variable model is a simple linear regression on ln(CO2) with a 14-year lag. I chose 14 years because that's the lag at which correlations maximize between ln(CO2) and GISS temperature; it physically represents the time for a CO2 pulse to warm the mixed layer.

    The point is that one can use this simple model to forecast global temps for 14 years in advance with a fair degree of confidence, based on current CO2 levels. That's why the red line extends forward from today. (I estimated 2015 and 2016 CO2 levels at +2 ppmv per year, but it won't be much different than that.)

    Looking closer, we can see that the initial period chosen was a bit unfortunate, since 2005 five-year running mean was a rather high point (compared to the model):
    https://farm8.staticflickr.com/7283/16841271636_fcf75d8cbf_o.jpg

    ... which means that the baseline of the bet is a little higher than perfectly average conditions would have meant. Nevertheless, the key points are found in the pink diamonds, which represent where the blue line would have to be at those times for money to change hands.

    Because of the terms of the bet, ELI WILL NOT LOSE MONEY unless the blue line falls below the LOWEST pink diamond at any of the three points. If the blue line passes above the top diamond, Eli wins $2000 for the period (wins both bets); if the blue line passes between teh top diamond and the middle diamond, Eli wins $1000 for the period (wins one bet, the other is a push); if the blue line passes between the second and third diamond, no money changes hands (Eli wins one and loses one); and only if the blue line passes below the lowest diamond does Eli lose (both bets, for a $3000 loss).

    I think you will agree that it is quite unlikely Eli will lose a dime; but even if he does lose in 2019, he will easily make up for it in the succeeding periods.

    ReplyDelete
  6. Jon

    I've iterated it four or more times this far, here, and if memory serves , Greg Laden's ATTP and Sou's,

    The idea is to have pros like Ladbrookes or Jimmy the Greek offer a bet sheet would be a simple decadal matrix of Delta T's from .1 to 1 decree C per decade.

    I'd expect no one to offer odds on the low end, and no one to bet for real on the high .

    Thus far neither side has rushed to bet on what its PR men are marketing ( or is framing the word ?)

    ReplyDelete
  7. Apropos of which, sort of, Martin Wolf in the Financial Times (not renowned for its liberal views):

    http://www.ft.com/cms/s/0/790b9350-6678-11e4-8bf6-00144feabdc0.html#axzz3Uv9r2vWV

    ReplyDelete
  8. Fergus, that's paywalled- what's Wolf saying ?

    ReplyDelete
  9. Thus far neither side has rushed to bet on what its PR men are marketing ( or is framing the word ?)

    How could we when Jimmy the Greek is dead and you haven't convinced Ladbrokes to offer the market you wish existed?

    My personal experience, back when Intrade offered climate markets, was that no one was willing to bet as though they really believed temperatures were sure to plunge anytime now, or at least climb no higher, no matter how ardently they claimed to believe that was sure to happen. Also, the markets in general were very thinly traded, especially the ones that wouldn't pay off for even a few years. Good luck getting very many people to tie up money in bets on decadal timescales, even if you somehow talk Ladbrokes into offering the market.

    ReplyDelete
  10. Russell, it's naughty to c&p articles, but the first 2 paragraph give a flavour:

    High quality global journalism requires investment. Please share this article with others using the link below, do not cut & paste the article. See our Ts&Cs and Copyright Policy for more detail. Email ftsales.support@ft.com to buy additional rights. http://www.ft.com/cms/s/0/790b9350-6678-11e4-8bf6-00144feabdc0.html#ixzz3UwLq6GEB

    T
    he Republican victory in the midterm elections was a triumph for its strategy of sustained vilification of the president and obstruction of his policies. The result will have big implications for the future of the US. But it also has implications for the rest of humanity. This is inevitable, given the role of the US as the world’s largest and most technologically advanced economy, guarantor of the open world economy and greatest military power. But the US is also the world’s second-largest emitter of greenhouse gases and among the highest emitters per head.
    The most important consequence of this election may therefore be to bury what little hope remained of getting to grips with the risk of dangerous climate change. Countries cannot keep bits of the atmosphere to themselves. Moving off the world’s current trajectory is a collective task. Without US will and technological resources, the needed shift will not happen. Other countries will not – indeed cannot – compensate.

    There's more good stuff in there...

    ReplyDelete
  11. Jon, you are the very first person to respond.
    Feel free to take it to whom you please, dead or alive

    Fergus, the problem is that the current trajectory
    is a line headed rather slowly uphill, not a visibly exponential curve , making it very hard for politicians to prefer entities as metaphysical as models and GCM intercomparisons over the linear trend

    ReplyDelete
  12. Russell,

    It appears likely to me that everyone else, of the limited number who have read your comments on the topic scattered here and there on other people's blogs, has correctly perceived that you aren't actually serious about wanting to bet with anyone, you just enjoy referencing people's unwillingness to put their money where their mouths are as a rhetorical device. If you're serious about wanting to bet, why don't you have a post describing exactly what kinds of bet you'd be interested in prominently on display at your own blog, VVatts Up With That?

    ReplyDelete
  13. " you just enjoy referencing people's unwillingness to put their money where their mouths are as a rhetorical device."

    Inaction speaks louder than words.

    ReplyDelete
  14. The Bunny's Munny, part 2.

    Using 5-year averages rather than annual data adds a little detail: since the 5-year average ending in (say) 2005 has its average year at 2003.5, the best correlation occurs not at 14 years lag from 2005, but at 17 years lag (which corresponds to a 14-year lag from 2003.5). So we have to draw a slightly different correlation. Here I also dispense with the year and just regress 5-year average GISS temperature anomaly against the 17-year lagged CO2:

    https://farm9.staticflickr.com/8734/16256795624_7a9829cd73_o.jpg

    Assuming that the errors are normally distributed (and they're not, because of the effects of autocorrelation) Eli has a 68% chance of winning all six bets. He has a 98.3% chance of winning all three bets at 2-to-1 odds. The expected value of the bets in 2019 is $1631 for Eli; in 2024 the expected value is $1935 for Eli; and in 2029 the expected value is $1987 for Eli. For the entire duration, the expected value is $5553 for Eli.

    ReplyDelete
  15. Forgot to mention that the graph above is only the years from 1950 forward, when the errors calm down considerably.

    ReplyDelete
  16. KAP, of course, illustrates the differences between the nominal level of statistical significance accepted for scientific journals and betting odds. Betting odds are what is used on the street, e.g. for policy. No one lives in a journal

    Conceding that statistical significance is needed for policy decisions is harmful and insane.

    ReplyDelete
  17. Inaction speaks louder than words.

    What action have you taken? All I see from you is words. Not even words on your own web site, where you could maximize the visibility of your desired bet if you wished?

    I bet actual money on climate on Intrade while it operated. I also made a commitment to the collective bet based on satellite temps for the current decade versus the last one on the No Tricks Zone web site (google "Climate Bet for Charity"). What actions have you taken that prove you're serious about betting on climate?

    ReplyDelete
  18. Some Corrections

    I have posted a fuller explanation and some corrections to these comments on Daily Kos.

    ReplyDelete
  19. Russell,

    >"If you're serious about wanting to bet, why don't you have a post describing exactly what kinds of bet you'd be interested in prominently on display at your own blog, VVatts Up With That?"

    Seems reasonable to me. More details are definitely needed if you want to make a bet (perhaps these were provided on another blog you can reference).

    Are you interested in a trivial bet? Just for shits and giggles. Im thinking, the bow tie in your profile picture for... something of equal value. You choose the terms, and others will take or leave.

    ReplyDelete

Dear Anonymous,

UPDATE: The spambots got clever so the verification is back. Apologies

Some of the regulars here are having trouble telling the anonymice apart. Please add some distinguishing name to your comment such as Mickey, Minnie, Mighty, or Fred.

You can stretch the comment box for more space

The management.