Discounting the future is all well and good, but basically accepting that people in two centuries will have to deal with ten metres and that’s not our problem is deeply immoral.

I must say that your information has been likewise fascinating to me!

Thanks in return

I’m not sure if this is the same as a decision tree technique, where you draw up the tree with the various decision nodes and costs, but you leave the probabilities flexible. First you set your best guess at all the probabilities, and see which set of decisions gives the best outcome.

Then you say, “What would the probability have to be, in order to tip the balance and make another set of decisions preferable?” So you start progressively changing the probabilities and see how far you have to go before your decision changes.

For a practical example, we could consider peak oil and buying a new family car. This makes an interesting tree.

The first decision node might be “Buy now, or wait a few years and keep driving the old gas guzzler?” If you wait a few years, then you put in a set of average petrol costs (e.g. $1/litre, $3/litre, $5/litre) and thus running costs for the old banger, plus higher maintenance costs for the aging car.

If you decide to buy now, then you put in the sale price on your old banger and go straight into another decision node “Buy small petrol, Buy diesel, Buy hybrid?”. For each alternative, you put a purchase price, and a set of running costs based on average petrol costs.

Then in say 5 years, plug-in electric and fully electric vehicles are available, so you have another decision node “Sell the old banger now, or continue running it into the ground for another 5 years?”. If you sell, then you put a sale price and a decision node for the different types of new vehicle.

For the other branches (where you bought a new car 5 years ago), you could also have a decision node for “Trade-in on a plug-in hybrid or fully electric?”

Meanwhile peak oil may have hit big time, so fuel prices may have skyrocketted and old bangers are only worth scrap metal.

Anyway, the upshot is that you can fiddle with the probabilities of different oil prices, and see at what point the decision flips from one best outcome to another.

To put it simply, if we have say 90% probability of today’s fuel prices continuing, then keeping the old banger is probably going to be the best financial decision. If we have a 90% probability of $5/litre oil, then the decision will probably be at least diesel, and possibly hybrid depending on purchase cost and maintenance costs.

Perhaps it tips over from old banger to diesel/hybrid at 50% probability of high prices? If this seems easily achievable (subjective assessment of probability) given the way things are going, then it tells you what to do, even though you don’t know the exact probabilities.

This would be how you can justify the subjective decision process on reasonable financial grounds.

But of course cars are not a purely financial decision. As better half has explained to me, they are also an emotional/image/performance decision for most men, which is why Top Gear is so popular…

If people want to think more about Antarctica I did a fairly detailed post on the Wilkins ice bridge collapse which tried to put it in some context of what was happening overall.

There is an interesting post by Rahmstorf and Vermeer at RealClimate. What’s going to happen in the 21st century? Too hard.

Stefan Rahmstorf’s comment at comment 33 is a worry:

I don’t think one can conclude a priori that future sea level rise cannot be faster than that during deglaciation.

We are pushing the climate harder than it’s been pushed probably since 65mya.

this stuff is precisely what I work on, so interesting post. All the tools you mention can be used in various contexts. It gets even hairier than you describe – there are multiple framings of risk – you won’t get a coherent description of the decision rules from policy makers. Those that do have a good grasp on it don’t want to telegraph what they have done to manage the political risks because that detracts from the other risks! One of my responses to these multiple framings and debates held at cross purposes is to get more deeply into cultural theory and the psychology of risk perception. As researchers, one of the thing we have to do is to hold a mirror up to these multiple views, so that people can gain a broader perspective.

Subjective probabilities in uncertain science can be tested for what impact they may have on decisions – this tells you whether the uncertainty matters or not. This technique has been under-utilised.

The incompletely quantified sea level projections by the IPCC in AR4 were an error of judgement. The climate modellers still think in forecasting terms – they also need to get a better grounding in managing uncertainty and risk using a combination of Bayesian methods and techniques used by the social sciences.

There’s an interesting set of classifications that could be made for risk management:

(a) Decisions where you can assign probabilities to the alternative outcomes (due to a lot of events which have provided historical data)

(b) Decisions where you have no real idea of the probabilities of alternative outcomes (because it has only happened a few times, or never happened before)

(c) Decisions where the outcomes from your decisions are affected by other’s decisions.

A lot of decision analysis and risk management theory has been developed for case (a), with the use of decision trees and monte carlo simulations and EMVs (Expected Monetary Value) weighting the outcomes according to their probabilities.

That’s all very nice and good. The bugger of it all, is that climate change is most likely case (b) with case (c) thrown in for good measure.

In short case (c) is the province of game theory – you know, zero sum games, prisoners dilemmas, tragedy of the commons, etc. This is the problem for Copenhagen – how to get everyone to participate and play fair.

The interesting thing here, is how do you assess case (b) for risk management?

The proposal is that you draw up a table, which identifies a set of choices/actions (rows), and a set of scenarios (columns) for High, Medium and Low conditions. Then you populate the table with your outcomes. You can’t assign probabilities reliably, but you can make an estimate of the cost of different outcomes.

Looking at your table, you can discuss what “decision rule” will apply:

1. maximax (maximise upside) – risk seeking

2. maximin (minimise downside) – risk aversion

3. minimax regret (minimise opportunity cost) – minimise the cost of backing the wrong horse.

For example, in the case of investment in late 2007 we were headed for a possible major calamity, or then again maybe a “stronger for longer, supercycle” depending on which set of economists you believed. Given the inability to predict and assign probabilities, we could have decided the leadup to the GFC was a case (b).

We could have made a simplistic decision table with 3 rows (all in stocks, 50-50 stocks and cash, all cash), and 3 columns (boom, plateau, bust).

Obviously you get the best answer for decision rule 1 with all stocks-boom, and rule 2 is all cash-bust. The classic optimist-pessimist contrast.

This is where Isaac Newton went wrong with investing in the South Sea bubble – first optimisism (huge stock holdings), then pessimism (sold out), then regretting (bought in again), then crash! Oh bugger…! He should have used Rule 3.

Rule 3 (minimax regret) is the interesting one – we could have acknowledged a priori that we would regret shifting to all cash (8% interest or so) if the 2007 boom continued, and more seriously regret all stocks if a 2008 bust occurred. Minimising your regret, or “minimising your opportunity costs” if you prefer that expression, leads to a 50-50 choice in this simplistic example.

So what has GFC to do with climate change? Not much except that CC is another, more difficult case (b). We will all be the wiser in retrospect, with our 20-20 hindsight. But we have to make investment decisions today (no action on CC is a form of defacto investment decision).

Can we Aussies, let alone the G20 countries, even agree on a range of outcomes and the relative costs in each case?

Which decision rule are we arguing for?

Some who are strongly convinced of worst case scenarios (e.g the Greens?) would probably argue for Rule 2 immediate strong action? Those who aren’t yet convinced of climate change (e.g. Barnaby Joyce and more than half of the Liberals and Nationals) may argue for Rule 1? There is little common ground here.

In between, could we have the ALP and the enlightened Liberals with some sort of Rule 3, hedging our bets or minimising the opportunity costs from a wrong decision?

Would it help break the deadlock on this upcoming debate, if the Aussie electorate and pollies were clear and agreed on what decision rule we are applying?