Be wealthy

Bret Stephens has a new op-ed in the New York Times about Hurricanes, Harvey and the Capitalist Offset. His conclusion is that the storm will be a speed bump to Houston’s economy and that

[t]he best lesson the world can take from Texas is to follow the path of its extraordinary economic growth on the way to environmental resilience.

Essentially, get wealthier and you will be more able to deal with environmental challenges.

The problem I have with this is not that it’s wrong (I’m sure it’s roughly true) but that it’s just trivial. It doesn’t really tell us anything about how to do so, and it ignores – in my view – many relevant factors. A response I particularly liked was this one, by Joseph Majkut, which points out that we should really be considering what it would cost to follow a pathway that reduces the impact of climate change and comparing that with how much of a reduction in damages that would produce.

However, rather than discussing this any further, I thought I would just make some points that I think are relevant and often ignored.

  • The issue is not about fossil fuels specifically, but about the emission of CO2 into the atmosphere. If we could find a way to use fossil fuels without emitting CO2 into the atmosphere, then that would be one viable solution. However, I do think that we should be careful of banking on technologies (such as negative emissions, or CCS) that have currently not been shown to work at the appropriate scale.
  • We have the potential to emit enough CO2 that the resulting impacts would almost certainly be regarded as catastrophic. Furthermore, even if we do not emit as much as we possibly can, we still can’t rule out that the impacts will still be very severe. Maybe we’ll be lucky and climate sensitivity will be on the low end and we end up not emitting enough to produce changes that are severe. Maybe we’re lucky and we end up reducing our emissions without really considering how to do so. However, assuming that this will be the case seems a poor way to address this; we can’t rule out the possibility that we could follow a pathway for which the impacts of climate change will be so severe that we will not simply be able grow ourselves out of trouble.
  • A key thing that I think is misunderstood is that CO2 accumulates in the atmosphere; it’s not about stabilising emissions, but getting about them to zero. As long as there is a net anthropogenic emission of CO2 into the atmosphere, it will accumulate. This means that as long as we are emitting CO2, we will continue to warm, our climate will continue to change, and the impacts will continue to get more severe. This implies that if we continue to rely on fossil fuels that emit CO2, then there has to be a level of growth below which climate damages will almost certainly grow relative to the size of the global economy.
  • About 20% to 30% of our emissions (depending on how much we emit in total) will remain in the atmosphere for thousands of years, and the resulting changes will effectively be locked in. This means that climate change is essentially irreversible on human timescales. If we do too little now, and the impacts turn out to be very severe, we don’t get to try again. Yes, maybe we can develop technology that removes CO2 from the atmosphere, or cools the planet by reflecting some extra sunlight back into space, but these will carry their own risks and will probably be technologically challenging. Certainly my view is that it would be better to think of ways to avoid possibly having to rely on technologies that we’ve yet to develop.

So, I certainly don’t dispute that becoming wealthier is an effective way to deal with environmental challenges. However, implying that we can grow wealthier without actually considering the potential environmental challenges, seems remarkably naive, especially as we do have the ability to produce changes to our climate that could lead to impacts that are extremely severe.

Links:
Trash arguments from Lukewarmer Oren Cass.
When integrity checking is more important than fact checking: More bad faith from Bret Stephens.
NY Times climate-misinformer tells Houston: You’ve ‘never been safer’.

Advertisements
This entry was posted in advocacy, Climate change, Climate sensitivity, Global warming and tagged , , , , , , . Bookmark the permalink.

139 Responses to Be wealthy

  1. jsam says:

    How much wealthier will Bangladesh have to become?

  2. Well, yes, Bangladesh would need to become much, much wealthier if it were to have the same resilience as Houston.

  3. JCH says:

    Houston has no resilience. In large part, they will not be paying for this. Americans will pay for it. It’s going to get interesting. The 21st-century tab, starting with TS Allison, is getting very very very big. Professor Curry’s 12-year drought did not exist in Houston. Houston has been pummeled with damaging storm after damaging storm.

  4. jacksmith4tx says:

    The Onion saw this coming…

    Authorities Urge Residents To Evacuate Dangerous Lower Income Brackets
    http://www.theonion.com/article/authorities-urge-louisiana-residents-evacuate-dang-56809
    “We know from experience that in hazardous conditions like these, the safest place for Louisianans to be is at least four or five times wealthier than they are now. This is no time to take risks—please, leave right now and make your way to the upper-middle class if at all possible.” Edwards went on to say that while no one could be forcibly evacuated, anyone who chose to remain in a lower income bracket should not expect to receive help anytime soon.

  5. There remain lots of ways a Houston can rebuild to create wealth which really matter in the long run, even setting aside considerations of climate change for the moment. For example, bailed out by a special funding package from Congress, they could choose to rebuild most everything where it was. That might get wealth creation on track, but it leaves the city susceptible to the next Hurricane Harvey or worse, even if these were not getting worse, which they appear to be. Call that short-sighted. Alternatively, a Houston could learn the lessons of a failure to have a city plan which is resilient against their frequent floods, and pursue a rebuild which is more robust over the long term. In the short term, such a plan might well track lower in wealth creation over time than the Let’s Put Everything Back. But in the long term, such a plan may well be better because there’s less periodic destruction.

    On the other hand, if the reconstruction is done using dollars which do not need to come from the Houston community, why shouldn’t they say Let’s Put Everything Back and forget any long term planning? They can bet on getting bailed out again if Badness Happens.

    I make this analogy because it also applies to climate change mitigation and adaptation planning. If the Rich Uncle is going to reimburse for everything bad that happens, there’s no cost to repeating the mistake and getting on the wealth wagon doing exactly what was done before. This is the moral hazard situation. Unfortunately, even at the federal level, the monies don’t come free, and, increasingly, as $25 billion per year or more is shelled out to reconstruct after storms, these monies are going to want to get offset by cuts in other programs.

    Politically, the problem is that no one wants to say to a Houston “You can’t have the money unless you rebuild differently” as wet, tired, impoverished people stand in front of TV cameras night after night.

    What’s going to be different about effects of climate change? Why won’t a catastrophe, say, a burst in sea level rise which finally does Miami Beach in, be addressed in the same manner? Why should Miami Beach fret? Won’t Congress pay?

    I think this is a big problem with preparing for climate impacts. Local officials bet that if it really is that big and bad, they don’t have to worry about it, because then it becomes somebody else’s problem. And why should people seriously cut back on CO2 emissions, unless, of course, they can make money doing it?

    No one should get me wrong: I think a Houston should be required to do the right thing, and that CO2 emissions should be required to be reduced. But that is is far from popular, not only with fossil fuel companies, but almost everyone around us.

  6. Andy Skuce says:

    I’m reminded of Homer Simpson’s remark that alcohol is both the solution to and the cause of life’s problems.

    So it goes with prosperity and growth. The benefits of prosperity for allowing cleaner and more resilient environment are unquestionable, but so also are the problems caused directly or indirectly by insufficiently restrained and regulated growth.

    Stretching the metaphor, a three-Martini buzz is fun but fleeting. But alcoholism and liver cirrhosis is forever.

  7. John Hartz says:

    Stephens’ coulmn reads like a synopsis of an Ayn Rand novel — pure fiction!.

  8. T-rev says:

    I can see this becoming the new troupe deniers branch swing to. That is, “Oh yes, alright, AGW exists but the solution is to triple down on more of the same, so everyone is rich. “, then we’ll waste another couple decades with this nonsense making it even harder to stop the mess we’re creating .

    Wealth is the exact opposite of what we need. Money is a proxy for energy expenditure, the richer you are the more you spend (bigger houses, business class flights or personal jets, more cars, more stuff etc) and now we have a whole swag of rich liberals thinking the only thing needed is a billion electric cars and few billion wind turbines to run them.. and apparently we do this without strip mining places like the Amazon for minerals.

    As to carbon capture, it’s not just the capture, it’s the sequestration, for millennia. If you capture and store for a decade or two and then it starts leaking…

    and as Kevin Anderson points out, the energy penalty in carbon capture is enormous

    All of that aside, Huston isn’t paying for this, the losses are socialised and the US taxpayer AND the world will, via buying US debt that can never be repaid and the crisis will get worse and worse until collapse is inevitable.

    https://tamino.wordpress.com/2016/12/10/climate-change-by-the-billion/

  9. Nick says:

    Clickbait Stephens scores hole-in-one at the mini-golf

  10. Steven Mosher says:

    “Politically, the problem is that no one wants to say to a Houston “You can’t have the money unless you rebuild differently” as wet, tired, impoverished people stand in front of TV cameras night after night”

    imagine letting communities pay for their own diastereous choices. putting the cost where the risk is. Stephans is right. it should be a road bump for houston because others who did not enjoy the growth will pay for the forseeable unintended consequences.
    end fema.

  11. Ragnaar says:

    Solomon M. Hsiang, Amir S. Jina

    You have a measure a wind speeds. Higher event speeds are worse. You find growth was less after some high wind speeds. How do you know what growth was supposed to be? How do you control for everything else? Most of the plots start in 1970. I’ll guess, the model uses 20 year blocks to say what growth is supposed to be for the next 20 years.

    What if most of these economies are transitioning from a growth phase to a mature phase? If a model tells me what growth was supposed to be, I don’t think I am buying it. Let’s ask a model to tell us the impact on our GDP of lower corporate tax rates.

    They find that 3.7 years of growth is lost over 20 years. Would you accept a 4% return instead of a 5% return?

    “Macroeconomics, he argues, is like a science that has not only stalled for three decades, but has actually gone backwards in its ability to understand reality.”

    “Romer’s huge mea culpa on behalf of mainstream economics is a sign that, after a decade-long hunt for trolls and gremlins as the cause of crisis, academia now has to begin the search for the cause of instablity inside the system, not outside it.”

    https://www.theguardian.com/commentisfree/2016/sep/19/its-time-to-junk-the-flawed-economic-models-that-make-the-world-a-dangerous-place

    I am not against GCMs. Economic models is a different story. It’s been more than 20 years since I learned the Keynesian economics were not going to work often enough. Would you day trade with an economic model? I am sure about twenty people will gladly sell you one.

  12. Everett F Sargent says:


    The best lesson the world can take from Texas is to follow the path of its extraordinary economic growth on the way to environmental resistance.

    Yeah, that’s the ticket. Resist the environment.

  13. JCH says:

    It’s impossible. The United States of America is stuck with Houston, and New Orleans, and Miami, etc. It’s also stuck with politicians who will declare national days of prayer, and the best ever rebuild, the best ever FEMA, etc. The country is determined to be stupid far past the foreseeable future.

    Fortunately, money is not real and the actual cost of rebuilding Houston is unknowable, but probably an economic benefit.

  14. @T-rev,

    I agree with a lot of your comment, but quibble with this:

    and apparently we do this without strip mining places like the Amazon for minerals.

    That’s an often-heard position, but it is simply incorrect. No one builds wind turbines with magnets in their generators any longer, the blades are increasingly Carbon composite, and the future chemistry for batteries, admittedly not available right now, is not Lithium or Manganese or Cobalt, but organic batteries, based upon Carbon. Many of the “rare earth” kind of materials were used for these because they were available and familiar, but cost pressures are pushing new developments, including how readily the stuff using these technologies can be rolled back into the process after their lifetimes.

  15. Willard says:

    BrettS is to ClimateBall what Juicero is to juice makers:

    One costs 400$. The other costs a bit more.

  16. Steven Mosher says:

    what JCH said

  17. Willard says:

  18. Willard says:

    You might also like:

  19. Everett F Sargent says:

    So, say you have home owners insurance, a necessary prequirement, for securing your 30-year mortgage. You had no flood insurance though, either because you could not afford it, or because you were not within the 100-year floodplain according to FEMA (which really shops out all that type of work to either the USGS or the USACE). Which type of bankruptcy do you apply for, Chapter 7 or Chapter 13?

    “Fortunately, money is not real and the actual cost of rebuilding Houston is unknowable, but probably an economic benefit.”

    Yeah, that’s the ticket! Bankers and realtors and contractors and insurance agencies stand to make much $$$, flatlanders, err homeowners, not so much.

    In fact, we should do this all the time, anthropogenic global wastage, because it will always be an economic benefit, for the weiners, at least. 😦 😦 😦 I just broke all my toys, because someone else suggested that I would gain an economic benefit. Why leave to nature, when humankind is so much better at breaking and taking stuff.

  20. JCH says:

    We do do it all the time. Christmas. Naughty kids still get toys. We live in a monetary fairytale. It’s not all that bad.

  21. Susan Anderson says:

    Stephens follows Pielke Jr. in this article, and is known to quote Lomborg. He’s a New York Times Opinion writer (with a Pulitzer) and hates Trump. The NYT has hired quite a few excellent climate science writers recently, and they’re making an effort to cover it from a variety of points of view. Unfortunately, Stephens has a devil that insists on pushing his unskeptical climate views on the world. You need only look at Readers’ Picks to see how unwelcome he is there on climate.

    I could wish that false accusations of supporting Trump and the teaparty were not made. False accusations have a way of putting people’s backs up, and I think without them he might be one of the few who would take a harder look at their assumptions. However, anyone clinging to Lomborg and PRJr is intellectually nude in public with most scientifically literate people.

  22. Marco says:

    As I wrote over at the Stoat’s den:

    Is it really so simple as blaming the Houstonians to live where they live, so let them suffer their own choice of living there without proper protection/insurance?

    If so, remember that there is a lot of industry there, on which the rest of the American economy depends. Katrina hit oil production hard, causing long-term increases in price. Harvey has also hit oil production (and derivatives) hard, and thus will also cause long-term increases in price. In fact, some have warned of a real shortage of ethylene, a major starting chemical for lots of different products, for months.

    The rest of the people, not taking the same risk as the Houstonians, may well find themselves wondering why they suddenly have to pay higher prices, even when not a single dollar is transferred to help the Houstonians recover…

    If you really want, you could say the externalities of having important production in Houston aren’t properly included in the price of the product, but are to be paid when disaster (in this case Harvey) strikes.

  23. There’s a symmetry noted in Oxfam’s still relevant Extreme Carbon Inequality report

    https://www.oxfam.org/sites/www.oxfam.org/files/file_attachments/mb-extreme-carbon-inequality-021215-en.pdf

    … While the richest 10% in the world are responsible for 50% of emissions,
    the poorest 50% in the world are responsible for 10% of emissions.

    The poorest have few or shit choices; the richest need to start making the moves (I’ll include myself in that).

  24. dikranmarsupial says:

    “Essentially, get wealthier and you will be more able to deal with environmental challenges.”

    The problem here is that it is only you that will be more able to deal with environmental challenges, but if you becoming wealthier means you contribute to environmental challenges to others or contribute to others becoming less wealthy (and hence less able to deal with environmental challenges) then you are behaving unethically (if you accept the golden rule, which seems a reasonable starting point to me).

    So either (i) we become wealthier without adding to environmental problems, (ii) we become (globally) a more equal society, or (iii) we have to accept that we are jerks that don’t meaningfully care about others. (i) is difficult (ii) is politically unacceptable for many and nobody really wants (iii), which is why we end up with denial and argue against climate science, even on topic where it is obviously very solid (e.g. cause of increased atmospheric CO2, second law of thermodynamics etc.), rather than getting to grips with the issues that really stop us from doing something about it.

    The idea is fine, but you have to work through the consequences, making us all wealthier (reative to the environmental challenges) would be great, but that is a lot more difficult than making already wealthy people a bit wealthier.

  25. Stephens’ is the Schelling Conjecture: Impacts fall faster if you invest in growth than in emission reduction.

  26. Andrew J Dodds says:

    Speaking as someone who is currently in hospital after yesterday’s chemotherapy session, I quite like living in a wealthy society; I’d be dead otherwise.

    I think that the error here is the automatic connection made between growth in energy use and growth in fossil fuel use. The obvious disconnect: coal and gas in electricity could have been replaced by nuclear power decades ago, and no one would have really noticed, growth-wise. It is a matter of choice.

  27. Stephens’ is the Schelling Conjecture: Impacts fall faster if you invest in growth than in emission reduction.

    Thanks, I hadn’t realised that there was an actual conjecture. I did find this by Anthoff & Tol, which says

    We use the integrated assessment model FUND to compute the income elasticities of climate change impacts for different world regions over time. We find limited support for Schelling’s Conjecture that development might be the best defence against climate change impacts

  28. Majkut’s website doesn’t seem to allow comments, but basically he outsources his entire argument to this quote:
    ‘But the size of that number has interesting implications. BNEF estimates that we only need an additional $5.2 trillion of investment to get on a 2°C path over the next couple of decades.’

    Which is complete nonsense. Investment in renewable energy since 2000 has already topped $4 trillion, comparable to the $5.2 trillion BNEF quotes when you account for inflation.
    http://euanmearns.com/worldwide-investment-in-renewable-energy-reaches-us-4-trillion-with-little-to-show-for-it/

    And the effect on emissions has been imperceptible. The decarbonization rate over 2000-2016 was actually lower than in previous periods. Even if the decarbonization rate increased by 1% a a year (from the historical 1-1.5% to 2-2.5%) it would only reduce century-end temperatures by 0.4 or 0.5ºC.
    https://wattsupwiththat.com/2016/12/30/do-the-math-climate-policies-cannot-change-century-end-temperature-more-than-0-5c/

    And yes, one should look at the decarbonization rate of GDP rather than simply emissions, because if you looked only at the latter you’d conclude Greece, Syria and Venezuela have done best in ‘fighting climate change’.

    So basically Majkut is saying the previous $5 trillion had no effect, or at least none that could be perceived in the global emission and GDP statistics. But don’t worry, just give us another $5 trillion and that will do the trick – you will go from ‘business as usual’ (whatever that means) to the 2ºC target just like that.

    (Actually, upon clicking the link I saw the BNEF estimate was only about electricity, which is the easiest sector to decarbonize – hell, it was already decarbonized in France, Norway and a bunch of other places with no need for climate conferences. Maybe Majkut should update his post)
    https://about.bnef.com/blog/mapping-the-gap-the-road-from-paris/

    Anyway, the question is not whether the economy is decarbozining – it’s been doing so since we started to put together statistics on the issue. The question is whether it’s decarbonizing faster in the times and places where massive investment and/or regulations to reduce emissions have been put in place. The answer is no.

  29. Hyperactive Hydrologist says:

    Do economic models include scenarios for mass global destabilisation caused by famine, mass migration or war as a result of climate change? What about the cost of managed retreat for coastal cities?

    Regarding the net zero CO2 emissions is it possible to make agriculture that will need to feed 9 billion plus people carbon neutral? And what about new infrastructure to support this population which will presumably requires materials that have an embedded GHG cost albeit lower than today.

  30. Albert,

    Which is complete nonsense. Investment in renewable energy since 2000 has already topped $4 trillion, comparable to the $5.2 trillion BNEF quotes when you account for inflation.

    Hold on, but isn’t that absolute, not relative? The relevant quantity is how much extra it would cost to follow a – for example – a 2oC pathway relative to some baseline pathway (presumably, one in which we don’t implement climate policies), not how much we end up investing – in total – in alternatives.

  31. JCH says:

    Impacts fall faster if you invest in growth than in emission reduction.

    And so the fairytale goes. One day we’re also going to reunite with the dead.

  32. Chubbs says:

    ATTP – Exactly. Incremental costs for renewable energy have dropped sharply since 2000. I’d argue the $4 trillion was a worthwhile investment. It hasn’t had a noticeable impact on economic growth and it has given us viable non-fossil alternatives. Global solar investment in 2017 will be higher than coal, gas and nuclear combined in 2017.

    http://www.saurenergy.com/solar-energy-articles/frost-and-sullivan-global-solar-power-investment

  33. Bernard J. says:

    Of all the environmental impacts humans have created, how many have been resolved by people becoming more and more wealthy?

  34. Andrew J Dodds says:

    Bernard J –

    In terms of acute pollution, the UK is much cleaner now than 50 years ago, and that’s because the surpluses arising from greater wealth were directed into being cleaner. The more wealth there is, the greater the political choices available, such as being cleaner. That we fail to make those choices is another problem..

  35. verytallguy says:

    Of all the environmental impacts humans have created, how many have been resolved by people becoming more and more wealthy?

    OK, I’ll bite.

    I was recently privileged to witness the flight of the California Condor in the Pinnacles National Park. Reduced to a couple of dozen birds, the wealth of the USA allowed a very expensive capitve breeding programme to pull back the species from extinction.

    Habitat management and protection, including at the Pinnacles, on reintroduction, now means a population of around 300 is in the wild. Every single bird is tagged.

    Only a wealthy population could have achieved this.

    That’s not to say I agree that growwwwwth will solve AGW. Far from it.

  36. Only a wealthy population could have achieved this.

    Indeed, being wealthier allows us to do things that may be of lower priority were we less wealthy. The problem, in my view, is that simply getting wealthier is neither all that simple, nor guaranteed. I certainly don’t disagree with the basic premise, but I don’t see how we can optimise our future pathway if we don’t also consider all the risks associated with the various options available to us.

  37. @Ragnaar,

    There’s modeling, and then there’s the failure to do the country level equivalent of zoning. What genius decided to build critical infrastructure like refineries — whether they are needed in the future or not — in a region which has a high probability of big storms and flooding? Ditto refineries and oil tanker docks in Philadelphia. That’s right no one did. It was left to whatever arbitrary amalgam of local, state, and federal tax and other policies made it convenient to do … a big ocean port, proximity to gasoline consuming centers, etc.

    I don’t think the price of anarchy has been properly reflected here, irrespective of other impacts. Why, I do not know. Whenever I get in a conversation about this, with someone knowledgeable about Economics, it seems to end up in some kind of circular reasoning that the price, by definition, is included in the going market price, to the degree it is important to everyone. Maybe, in fact, that’s not circular, but it seems so to me. Also, as others have mentioned in this conversation, where’s the counterfactual so you can really tell?

    And I’m perfectly aware that, in my nearby case of Boston, and for instances which I have testified in front of committees and commissions, despite a lot of good sounding words coming from Mayor Marty Walsh and some others, development of the shoreline continues, there’s increasing amounts of noise about building a big flood wall (which is the most ridiculous proposal other than the present plan — mass evacuation — which I can imagine), and this does nothing for other parts of the coast and areas like Dorchester which will, essentially, become a marsh (again) in the case of a big flood.

    And, like her counterpart, Senator Chuck Schumer in New York, Senator Elizabeth Warren rushes to the coast of Scituate and elsewhere to criticize the hardships imposed by higher flood insurance rates.

    Houston is just the latest instance. There’s a lot of this nonsense to go around. I’m almost to the point of despairing we’ll ever get this right, that the only way people are going to “get it” is when the bank breaks, or if their just constructed flood wall, billions of dollars having been spent on it, is made irrelevant within 10 years of its construction.

  38. Willard says:

    Being wealthy is good. Getting schwifty is better:

  39. Willard says:

    > Investment in renewable energy since 2000 has already topped $4 trillion,

    To put that number in perspective, Occupy Irak alone costed the American taxpayers (therefore excluding teh Donald who pay no taxes) more than three trillions:

    https://en.wikipedia.org/wiki/Financial_cost_of_the_Iraq_War

    And counting.

  40. I looked up a few things about energy usage.

    According to this, global electricity generation in 2014 was 23816TWh = 23816 million MWh.

    According to this, the average cost of electricity is around 15 US cents per kWh (or $150 per MWh).

    If the above is about right, then we spent $150 x 23816 million = $3.5 trillion in 2014 on electricity.

  41. @Alberto Zaragoza Comendador,

    I agree that there is little evidence the world has strayed much below the Business As Usual trajectory and that, still, in all rational forward-looking projections, RCP 8.5 is the way to go. Agreed, electrification of everything is the way to go, but it’s being done no where near fast enough. I don’t know what the clearing rate of Carbon intensive-dependent economic processes are in societies, including manufacturing, but just looking at automobiles as a relatively fungible case and using that as a lower bound, it’s gotta be several decades. I wonder if there’s a notion of half life there?

    Sure, fossil fuel companies and Trump and Pruitt and Koch and the gang all are factors. But I continue to believe that the average economically-well-off citizen of the United States is loathe to do thing which might impact their personal economy, including perceived home prices. And, unfortunately, there’s evidence that not-so-well-off citizens of the United States are too busy making it economically through the week to have any spare capacity to deal with all this. Klein’s This Changes Everything might apply, but not in the way suggested.

    And even for the countries which do decarbonize, it would be highly useful to underscore how much of that Carbon intensity they have exported to others. In Massachusetts, for instance, the state administrations under both parties thought they were doing pretty well under the local cap-and-trade program, until the Global Warming Solutions Act (GWSA) made them face up to the face that, actually, they weren’t, and buying Carbon credits from Vermont and New Hampshire is not the same as decarbonizing things locally. I think New York State had a similar awakening in their upstate/downstate emissions balance.

    The sober attitude is captured in a few recent mentions by Dr James Hansen and something which Prof Kevin Anderson has repeatedly underscored about UNFCCC and COP21, that any reasonable decarbonization paths in their proposals demand Carbon capture and sequestration, at unknown scale and costs. And I don’t even want to start talking about SRM! It is encouraging, though, that Harvard’s David Keith has jumped onto the sequestration path in a big way.

  42. @Willard,

    My son mused that if solar + storage has the leverage it appears to have, rather than people contributing to Sierra Club, Greenpeace, and the like, it seems more useful to create non-profits which simply erect solar PV coupled with storage everywhere they can, for zero cost to the occupants. There is such an organization in California, GRID Alternatives, but I don’t know how widespread the idea is.

  43. John Hartz says:

    You can bet your sweet bippy that the public health consequences of Hurricane Harvey article will not be resolved by the Ayn Rand-styled capitalism espoused by Bret Stephens and his ilk.

    From water contamination to diseases to mold in the walls, dangers continue long after the hurricane. These sorts of risks accompany climate change.

    Harvey Aftermath: A Public Health Crisis in the Making by Nicholas Kusnetz, InsideClimate News, Aug 31, 2017

  44. Mitch says:

    One thing that Stephens missed in his article is that the costs and benefits for rebuilding Houston are going to different people. The costs are going to homeowners, and if they have much of their wealth tied up in their home, their total wealth decreases. Even if the home is built back with federal funds, the likelihood is that home prices in a flooded neighborhood go down.

    The benefits go to the contractors and the building supply companies, so net effect is probably to increase income inequality.

  45. >> “Essentially, get wealthier and you will be more able to deal with environmental challenges.”

    > The problem here is that it is only you that will be more able to deal with environmental challenges, but if you becoming wealthier means you contribute to environmental challenges to others or contribute to others becoming less wealthy (and hence less able to deal with environmental challenges) then you are behaving unethically (if you accept the golden rule, which seems a reasonable starting point to me).

    > So either (i) we become wealthier without adding to environmental problems, (ii) we become (globally) a more equal society, or (iii) we have to accept that we are jerks that don’t meaningfully care about others.

    It isn’t clear this is true. Imagine a world in which (a) everyone gets wealthier, but adds to environmental problems; (b) incomes / wealth rise in proportion to current wealth, so society becomes less equal; (c) potential for damages increase in proportion to sqrt(wealth) but immunity-to-damage increases in proportion to wealth. In this version of the world, damages decrease, we all get wealthier, and we become less equal.

    I don’t mean to propose that the functions are this shape; merely that they could be; and therefore, that your “logic” is broken.

  46. Willard says:

    If an economic framework leads us to conclude that a natural disaster is beneficial, I suggest we revise our framework.

  47. Willard says:

    ­> In this version of the world, damages decrease, we all get wealthier, and we become less equal.

    Therefore (c) could very well apply: “we have to accept that we are jerks that don’t meaningfully care about others.” Which goes on to show that our Stoatness is yet again hard of reading.

    Freedom Fighters never seem to get to Branko’s last slide:

  48. Willard says:

    > Stephens’ is the Schelling Conjecture

    Then getting schwifty echoes Schelling’s theorem:

    [Y]ou’re standing at the edge of a cliff, chained by the ankle to someone else. You’ll be released, and one of you will get a large prize, as soon as the other gives in. How do you persuade the other guy to give in, when the only method at your disposal – threatening to push him off the cliff – would doom you both? Answer: You start dancing, closer and closer to the edge. That way, you don’t have to convince him that you would do something totally irrational: plunge him and yourself off the cliff. You just have to convince him that you are prepared to take a higher risk than he is of accidentally falling off the cliff. If you can do that, you win.

    I prefer theorems to conjectures.

  49. @ATTP: Indeed it turns out the $5.2 trillion quoted by Majkut is an incremental figure; the actual investment BNEF estimates would be necessary over the next 25 years is $12.1 trillion, but most of that would be necessary without decarbonization anyway (according to BNEF).

    Still, when accounting for a roughly 20-year difference between both periods (2000-2016 vs 2017-2042), assuming inflation of 2%, one has to bump up the original investment figure of $4 trillion by 50% to make it comparable to the latter period. So it’s really like doubling the investment, from about $6 trillion to $12.1 trillion. In both cases the figure is for the power grid only, excluding things like subsidies for heat pumps, electric vehicles, etc.

    The idea that this can cover the difference between ‘business as usual’ and a 2ºC pathway doesn’t make sense. The ‘business as usual’ used by the Hsiang paper Majkut speaks about is the A1B scenario, under which century-end temperatures are 2.8ºC higher than the 1980-1999 baseline.
    https://www.ipcc.ch/publications_and_data/ar4/wg1/en/ch10s10-5-4-6.html
    As that baseline is in turn about 0.7ºC above the ‘preindustrial’ 1850-1900 average, it’s really like a 3.5ºC scenario.

    Under a 2ºC scenario, century-end temperatures actually may have to be slightly below a 2ºC increase (versus an 1850-1900 baseline), to account for warming in the pipeline released after 2100. But even if one assumes that a 2ºC scenario actually means 2ºC increase by the end of the century (from 1850-1900), the difference with the A1B scenario is 1.5ºC. So basically Majkut is saying the $6 trillion invested in 2000-2016 had no effect, but investing anothe $12 trillion will make century-end temperatures 1.5ºC cooler than otherwise.

    For the rest of the century, warming depends basically on TCR and how much we emit; the ‘warming in the pipeline’ is only 0.7w/m2, most of it won’t get released this century, and the *difference* in how much of it gets released (depending on how much we emit from now until 2100) is even smaller.

    Nobody knows the exact value of TCR per doubling of CO2, but for simplicity one could assume it’s 1.5ºC (1.8ºC in the models, 1.35ºC in most datasets). Thus what Majkut is saying is that investing $12 trillion would make CO2 concentrations, by the end of the century, *half* what they would otherwise be. This is absurd; as an example it would mean ending up at 500ppm (impossibly low) instead of 1000ppm (impossibly high).

    Even more to the point, as we’re already at 410ppm, it would imply that cumulative emissions in the next 83 years will be several times smaller under the $12 trillion scenario than under the do-nothing scenario (increase of 90ppm versus 590ppm). That is impossible, at least if one assumes GDP would be the same size under both scenarios. Since Majkut’s and Hsiang’s very concern seems to be the economic damages of climate change, I assume they want the do-something GDP to be at least as big as in the do-nothing scenario.

    For a 1% increase in the decarbonization rate, cumulative CO2 emissions the rest of the century would be about 40% lower than if the decarbonization rate stayed the same. Not 80 or 90% lower.

  50. JCH says:

    Boom town; flood town?

    NATURE ALWAYS PREVAILS

    Since the District’s creation, and despite a history of successful flood damage reduction projects and progress throughout Harris County, close to 30 damaging floods have occurred in the area, resulting in hundreds of millions of dollars in damages in just under 70 years. However, after the 1940’s, the Harris County area did not suffer what would be considered a widespread, regional flood, that is, until June 2001.

    ALLISON: A NAME WE’LL NEVER FORGET

    Tropical Storm Allison suddenly formed 80 miles off the coast of Galveston, Texas, on Tuesday, June 5, 2001, no one expected that, five days later, it would go on record as one of the most devastating rain events in the history of the United States. Neither historical data nor weather forecasts could adequately predict this extraordinary storm that, before leaving the area, would dump as much as 80 percent of the area’s average annual rainfall over much of Harris County, simultaneously affecting more than 2 million people. When the rains finally eased, Allison had left Harris County, Texas, with 22 fatalities, 95,000 damaged automobiles and trucks, 73,000 damaged residences, 30,000 stranded residents in shelters, and over $5 billion in property damage in its wake. Leaving 31 counties with declared disasters in Texas, Allison went on to spread disaster declarations to Louisiana (25 parishes), Florida (nine counties), Mississippi (5 counties) and Pennsylvania (2 counties). Allison was the costliest tropical storm in the history of the United States.

  51. Willard says:

    While I still can’t find to what Richie’s referring, I added “any action combating global warming will be, intended or not, a foreign aid program” to the Contrarian Matrix:

    https://contrarianmatrix.wordpress.com/do-no-harm/

    and my source is this editorial:

    https://www.foreignaffairs.com/articles/1997-11-01/cost-combating-global-warming

    I added Richie’s name to the Colophon.

    Thanks!

  52. Willard says:

    For Rachel:

  53. dikranmarsupial says:

    @WilliamConnelly wirtes “”It isn’t clear this is true. Imagine a world in which (a) everyone gets wealthier, but adds to environmental problems; (b) incomes / wealth rise in proportion to current wealth, so society becomes less equal; (c) potential for damages increase in proportion to sqrt(wealth) but immunity-to-damage increases in proportion to wealth. “

    I think you are over-thinking it a bit, the point I was trying to make is wealth for individuals/communities/countries is not an answer to the global problem, however I’ll amend it to

    “So either (i) we become wealthier without adding to environmental problems, (ii) we become (globally) a more equal society, (iii) we have to accept that we are jerks that don’t meaningfully care about others or (iv) a miracle happens. “

    (i) (ii) and (iii) are, I think, objectively do-able. It is not clear to me that (a) is possible in a world that is not resource-rich with respect to its population size, but I am no economist. I’m not saying there is a satisfactory solution, but I think all of the reasonable compromises are going to involve collective action, rather than us each solving our parochial problems individually (e.g. by getting wealthy).

  54. > I’ll amend it to… (iv) a miracle happens.

    No; that’s not good enough. I could say why, but if you can’t think of it for yourself there’s no point.

  55. Ragnaar says:

    hypergeometric:

    I guess there is a pricing failure. The sea side is risk and reward. Same with areas like Houston and barrier islands. Houston as another mentioned is part of our energy infrastructure. We’ll always have economic activity near the oceans. I suppose ocean shipping is the same as always. Rewards and risks. Look at what’s happening in Pacifica. People are drawn to the seas.

  56. John Hartz says:

    Another antidote to Stephens’ fairy tale…

    With Enormous Harvey Damage, Naomi Klein Warns Against ‘Disaster Capitalism’ Redux

    Now is exactly the time to speak out against free-market exploitation, says author of ‘The Shock Doctrine’

    byJulia Conley, Common Dreams, Aug 28, 2017

  57. “Imagine a world” where people use ludicrous hypotheses in order to arrogantly assert they have won an argument but have instead said more about themselves than it is polite to record here.

  58. John Hartz says:

    As documented in the article below, Bret Stephens’ core message flies in the face of Einstein’s admonition, We cannot solve our problems with the same thinking we used when we created them.

    Houston’s flooding shows what happens when you ignore science and let developers run rampant by Ana Campoy & David Yanofsky, Quartz, Aug 29, 2017

  59. John Hartz says:

    A take-away paragraph from the Quartz article I cited above:

    The Harvey-wrought devastation is just the latest example of the consequences of Houston’s gung-ho approach to development. The city, the largest in the US with no zoning laws, is a case study in limiting government regulations and favoring growth—often at the expense of the environment. As water swamps many of its neighborhoods, it’s now also a cautionary tale of sidelining science and plain common sense. Given the Trump administration’s assault on environmental protections, it’s one that Americans elsewhere should pay attention to.

  60. @wotts
    Limited empirical support means, of course, that there is empirical support, albeit in a few cases only.

    If you compare and contrast the death toll of recent floods in the USA, India and Niger, you see that the Schelling Conjecture is true in this case.

    Material damage is more subtle, but a number of papers have appeared over the last ten years that show that, proportionally, the damage in the long-term is much larger for poorer countries.

  61. Richard,

    Limited empirical support means, of course, that there is empirical support, albeit in a few cases only.

    Yes, of course.

  62. Chubbs says:

    There just aren’t enough fossil fuels to give the world a Houston lifestyle.

  63. Willard says:

    I don’t think the choice is to turn the industrialized world into Bangladesh, Richie, so I doubt Schelling conjecture has been tested much, if at all.

  64. JCH says:

    Hurricane Harley hit in a relatively low population area: Aransas Pass (8,500 people) and Rockport (11,000.) There were a lot of evacuations before the storm hit.

    Houston, which is inland, was not hit with a hurricane; it was hit by a tropical storm.

    The events are not that comparable.

  65. dikranmarsupial says:

    William Connelly yawn. If you want to explain why what you think I wrote is wrong (as I did for you) then I am listening.

  66. Steven Mosher says:

    When I want to understand economics I generally listen to economists.

    i liked this
    https://www.newton.ac.uk/seminar/20101209093010001

    https://www.newton.ac.uk/files/seminar/20101209093010001-152653.pdf

  67. verytallguy says:

    When I want to understand economics I generally listen to economists.

    Sound advice. I find that there is sufficient diversity of opinion amongst them that it’s always easy to find plenty that have the right understanding, i.e. they agree with my prejudices 😉

  68. dikranmarsupial says:

    SM interesting slides (as far as I understand it, which isn’t very far, perhaps the video might help, but I suspect I am a bit out of my depth), however the problem I am concerned with seems to be left to the “FURTHER ANALYSIS – extend the model to an n-country model”, as an earlier slide says: “Such research reflects a potential conflict of interests among countries and may have essential policy implications. “. “a country protects itself with adaptation” [emphasis mine] is the problem, some countries are unlikely to have the resources to protect themselves (by how much would Bangladesh’s economic wealth have to increase to adapt to the projected increase in sea level, not just for the next century but beyond?) and will need [considerable] help to do so, which is likely to be politically unpalatable in much of the developed world. It is also a matter of justice, why should a country have to protect itself against environmental damage that largely results from the actions of other countries? The real issues are not to do with the science, or really (although to a lesser extent) with the economics, but are social and political.

    ATTP-Style ending ;o) I don’t know a great deal about economics, but am willing to listen to anyone willing to engage constructively with what I have written.

  69. Marco says:

    Thanks, Steven. I opened the pdf and am now blind.

  70. Bernard J. says:

    As an Australian I am most familiar with the costs of environmental here. A very topical example at present is that of the Great Barrier Reef, which was recently valued at $56 billion, although that’s just the economic worth – the cost of actually building a Great Barrier Reef from scratch, including all of its attendant biodiveristy, is likely to be somewhat higher:
    https://theconversation.com/whats-the-economic-value-of-the-great-barrier-reef-its-priceless-80061
    And rebuilding it is exactly what we’ll need to do if we want to be around by the end of the century, because we’re on track to knock off 90% of the planet’s reefs by then:
    http://www.abc.net.au/austory/content/2016/s4624562.htm
    of course, the cost is not just that of the reef, because there’s the small issue of the acidity and the temperature of all of the globe’s oceans…
    Staying in the marine environment, Australia has pretty much lost its east coast Tasmanian giant kelp forests, and has started making inroads on the west Australian kelp forests. ANother ecosystem that’s pricey to reconstruct. Our alpine ecosystems will be largely gone by mid-late century, and our nothofagus associations along the spine of the Great Dividing Range will have been pushed uphill toward to extinction by the encroachment of schelophyll forest much more resilient to the increasing wildfires that we’re experiencing. Oh, and on that note, we can probably say goodbye to the cool temperate rainforest of Tasmania’s Southwest wilderness which, as I noted at Brere Eli’s a month ago, ois:

    …a World [H]eritage Area (look it up…) which contains a biologically significant and extraordinary remnant of Gondwanan flora and fauna relic species, [and] is already noticably suffering greater summer drought, leading to increased vulnerability to wildfire. I predicted about 5 years ago that by around mid-century the amount of summer drying, caused by global warming, would be such that large portions of the WHA would be so heavily burned that the cool temperate rainforest would be replaced by sclerophyl forest and more buttongrass plains. I was a bit optimistic with that projection – last summer saw serious wildfires that devastated significant areas of unique Gondwanan flora communities. The process of change of the cool temperate SW rainforest has already begun.

    Of course, it’s not all directly due to climate change. In Australia, and sticking with hip-pocket stuff, salinity is costing us somewhere between $200 and $700 million per annum, and acid sulphate toxicity somewhere around a $billion per year.
    I could go on and on listing the costs of our country’s environmental damage, but I suspect that the point is made. We’re supposed to be one of the wealthiest countries in the world, per capita, but our wealth doesn’t seem to be driving the state of our environment in the direction that it needs to go:
    https://soe.environment.gov.au/theme/overview/framework/state-and-trends
    The 2016 Australian SoE report (released this year) is heavily couched in bureaucratic weaseling and jargon, in order to drive it past the sensibilities of our current rabidly conservative government, but if one peels off the waffle the underlying message is grim. The message seems to be that the wealth of the nation is irrelevant if there’s no enlightened leadership to engage it where it’s needed.
    That’s a not-inconsiderable problem.

  71. Bernard J. says:

    “…the costs of environmental damage here…”

  72. Steven Mosher says:

    Marco.. best to watch the video.

    Maybe richie can come and explain the maths..

    One thing I appreciate is how guys like gavin, attp, and held, and rayP are able to
    make the complicated physics understandable, how they take a subject that is absolutely
    out of my grasp and bring it close enough that I get the idea, even if I never could have done it from scratch myself.

    Sure would be nice if economics dudes did the same thing..

    I’d sit quietly and listen

  73. dikranmarsupial says:

    “One thing I appreciate is how guys like gavin, attp, and held, and rayP are able to
    make the complicated physics understandable, how they take a subject that is absolutely
    out of my grasp and bring it close enough that I get the idea, even if I never could have done it from scratch myself.”

    I’d like to second that, especially given the other thread; even if he didn’t engage with the skeptics directly ATTP is doing a very good job of bringing the physics within range. Gavin, Isaac and Ray are pretty good as well! ;o)

  74. Marco says:

    “Marco.. best to watch the video.”

    They still show the slides, and the colours…the colours…the COLOURS…are screaming at me, especially on the background she used.

  75. JCH says:

    Too soon to know, but the Governor of Texas is now estimating damages up to 180 billion dollars. Even in oil-rich Houston, that is not pocket change.

  76. Isaac Held has stopped posting to his blog. Last year, I tried commenting with something that was clearly relevant, yet he rejected explaining “I did not accept your comment on post #70 because it was not on topic. The topic of the post is not the Coriolis force, with which most readers of my blog will be familiar.”

    Evidently climate scientists know how everything works. That’s why they are able to predict climate behaviors such as ENSO and QBO with exacting precision.

  77. Willard says:

    > https://www.newton.ac.uk/files/seminar/20101209093010001-152653.pdf

    TL;DR – Too Colourful, Didn’t Read. Except:

    Natali Hritonenko
    Department of Mathematics, Prairie View A&M University, TX USA

    Yuri Yatsenko
    School of Business, Houston Baptist University, TX USA

    Thierry Bréchet
    Center for Operations Research and Econometric

    No economist seems to have been harmed during the making of this presentation.

  78. Steven,

    Sure would be nice if economics dudes did the same thing..

    I’d sit quietly and listen

    Not economics specifically, but I do get the impression that there are fewer well known social science communicators, than science communicators – history would be one exception. I wonder if it isn’t partly because the physical sciences works on a framework that is well accepted. When we teach physics we focus initially on things like Newton’s Laws, Kepler’s Law, conservation laws and even something like the greenhouse effect can be explained on the basis of the fundamental concepts. I’m not aware of a similar framework within the social sciences, which may mean that it’s that much harder to lay things out in a way that is fundamental.

  79. Bob Loblaw says:

    “Even in oil-rich Houston, that is not pocket change.”

    I am willing to stand corrected, but I suspect that most of the $180b you mention will not be paid by people in Houston. This is a socialized cost across the entire U.S.

  80. Bob Loblaw says:

    ATTP: “I’m not aware of a similar framework within the social sciences”

    Well, having known a couple of economists for most of my life (father, oldest brother), and having taken two introductory economics courses (one in high school, one in university), there are basics such as “supply and demand”, business costs, etc. that are pretty fundamental and help me discuss economics with the better-educated. I’d agree that even these “laws” are a little more lax than fundamental physics, though.

    A key difference is micro-economics, which I think is fairly well defined in an empirical observation way, and macro-economics, which allows for a lot more hand-waving. Also keep in mind that there is a difference between a successful academic career and doing good work in economics. Of course, this is true in any discipline, but we know of at least one person (“economist”) that posts here that never seems to let gremlins get in the way of a successful career.

  81. JCH says:

    The annual Harris County budget is around 2 billion. Houston is not resilient. Texas is a storm damage welfare state.

  82. Nordhaus is out with a new damage estimate. An 8% drop in GDP from 6 °C global warming. That seems awfully optimistic. Our friend Richard gets mentioned right off the bat in the abstract.

    A Survey of Global Impacts of Climate Change: Replication, Survey Methods, and a Statistical Analysis, William D. Nordhaus, Andrew Moffat, NBER Working Paper No. 23646, Issued in August 2017

    The present study has two objectives. The first is a review of studies that estimate the global economic impacts of climate change using a systematic research synthesis (SRS). In this review, we attempt to replicate the impact estimates provided by Tol (2009, 2014) and find a large number of errors and estimates that could not be replicated. The study provides revised estimates for a total of 36 usable estimates from 27 studies. A second part of the study performs a statistical analysis. While the different specifications provide alternative estimates of the damage function, there were no large discrepancies among specifications. The preferred regression is the median, quadratic, weighted regression. The data here omit several important potential damages, which we estimate to add 25% to the quantified damages. With this addition, the estimated impact is -2.04 (± 2.21) % of income at 3 °C warming and -8.06 (± 2.43) % of income at 6 °C warming. We also considered the likelihood of thresholds or sharp convexities in the damage function and found no evidence from the damage estimates of a sharp discontinuity or high convexity.

  83. Steven Mosher says:

    ya willard I struggled with what word to use so I just looked at their publications

    https://scholar.google.com/citations?user=Wc-STbcAAAAJ&hl=en

    https://scholar.google.com/citations?user=i1D_TzsAAAAJ&hl=fr

    whaddah want call that kinda work? I’d be happy to have you edit my post with the correct non objectional noun that satisfies your essentialist proclivities.

    please feel free to correct.

  84. Steven Mosher says:

    ” I’m not aware of a similar framework within the social sciences, which may mean that it’s that much harder to lay things out in a way that is fundamental.”

    Metaphorically physics has a few canopeners,
    other fields tend to have dozens

  85. @Steven Mosher,

    In regarding your comment @Willard, with the Google Scholar links, looks like applied operations research to me. I’m not even sure “operations research” is a Thing, these days, but it certainly looks like applied optimization theory and methods. A lot of those people have jumped the energy barrier over into Machine Learning.

    I still run into them, though, doing Statistics. They are a lot like the ML people: With them, Algorithms rule, and the more twiddle factors, the better. I, as a Stodgy Old Bayesian (for the most part), strongly dislike free variables. I keep wanting to put priors on ’em.

  86. Willard says:

    > please feel free to correct.

    How about: “When I want to understand economics I generally listen to those who contribute to economics”?

    This way your focus remains on the product.

  87. Regarding twiddle factors: They are most successful when you know the behavior is deterministic. For example, conventional ocean tidal analysis can have hundreds of free parameters corresponding to all the possible lunar harmonics. And it works out pretty well despite the number of variables because the moon’s orbit is deterministic.

    So the challenge with something like ENSO is to see how far you can go assuming the behavior is deterministic and driven by external forces analogous to lunar tides. The stochastic part is the nuisance, as ENSO measurements do have noise.

  88. Marco says:

    “…and find a large number of errors and estimates that could not be replicated…”

    Gremlins, Bill. You found gremlins.

  89. Steven Mosher says:

    That’s an excellent suggestion! willard

  90. Steven Mosher says:

    “Nordhaus is out with a new damage estimate. An 8% drop in GDP from 6 °C global warming. ”

    They say Income.. I’m reading the paper.

  91. Steven Mosher says:

    “I’m not even sure “operations research” is a Thing, these days, but it certainly looks like applied optimization theory and methods.”

    My first title at Northrop was “Operations Research”

    Lots of theory and models, very little data.

    Example. I think you guys will enjoy this.

    The first “model” I worked on was a simple engagement model.

    two aircraft: you have to model the probability that one aircraft will kill the other.

    Major Events: Detection: Lock: Launch: Kill or Miss. At the event level you basically end up
    generating a tree of outcomes ( based on who detects first, who locks first, who launches first
    how many missiles you launch once you get a lock, and the Pkill of your missile and the number
    of missiles you carry. Super basic stuff.

    How do you choose the probability of Kill? especially if the missile has never been built,
    or if it has been built but only test fired? So you think how about historical data to bound
    extremes and do sensitivity testing.. Go look at missiles fired in actual combat

    talk about sketchy data. its a dogs breakfast
    https://defenseissues.net/2013/06/15/air-to-air-weapons-effectiveness/

    Operations Research is a thing. its not a pretty thing

    in business its a different thing, maybe related

  92. Steven Mosher says:

    “How do you choose the probability of Kill? especially if the missile has never been built,
    or if it has been built but only test fired? So you think how about historical data to bound
    extremes and do sensitivity testing.. ”

    then someone gets the bright idea to just model the missile. Its flight, its sensor, its navigation,
    the battery, the probablistic detection, the counter measures, the counter counter measures..

    And in the end… the missile is designed to explode a certain range from the target based on the proximity fuse

    Say its 30 Meters. The missile is travelling in excess of mach2. Maybe your frame rate is 30
    milliseconds.

    Opps;

    So now you got to end game… whats your Pkill?

    Oh shit.. Now you have to model overpressure, and the other modes of kill for the weapon
    ( exapanding rods and shit cutting through the airplane). How do you model that?
    With a structural model of the aircraft and some FAE to see what exactly gets damaged..?
    some clown asks for probability of damage. Some clown wants to know which systems
    are most likely to be damaged.

    So suppose you get all the way to the end.. what do you have? you still have a Pkill.
    But instead of one can opener ( assume the PK is .5) you have a hundred can openers
    and a pk of .48

    Clearly everyone agrees the higher fidelity model is better, but in the end you look at the
    messy real world and understand that other shit (like human behavior ) appears to matter more.

    So someone suggests that you put real men in the “modelled” aircraft and run mock battles
    on computers..90 million dollars worth

    Everyone likes those answers better. generals nod.

    And then the pilots say mock battles are nothing like real battles.

    and the argument over ‘best tool we have” rages

    and so it goes.

  93. dikranmarsupial says:

    hyper wrote “I still run into them, though, doing Statistics. They are a lot like the ML people: With them, Algorithms rule, and the more twiddle factors*, the better. “

    now now, we’ll have less of that, if you please! ;o)

    * over-parameterisation is not necessarily a problem if you have a prior over the function of a model, rather than over its parameters. We also try to minimise the number of “Twiddle factors” (i.e. things not directly inferred from the data by the algorithm), which is where a lot of the problems with inference lie, and putting a prior over them and marginalising is a sensible approach (IMHO).

  94. dikranmarsupial says:

    geoenergymath says: “Regarding twiddle factors: They are most successful when you know the behavior is deterministic.”

    AIUI everything above the quantum level (and maybe even there) is deterministic, there is no such thing as randomness (only uncertainty).

    Sometimes you need a model with lots of parameters so that it can describe the physics, but the more parameters you have, the more data you need to establish their values reliably (and it usually grows much faster than linearly in the number of parameters). Sure, use a model with lots of parameters, provided you can show that their values are well determined, that means you can say “X can be explained by Y”, to say “X is explained by Y”, you still need a plausible physical (rather than statistical) model. Overfitting is easier the more parameters you have and the less data you have, and the weaker you prior information is. Of course you can always reduce the amount of data necessary by having a strong prior, but then the results are only as good as your prior.

  95. Raypierre says:

    If as Brit Stephens says, this is just a speed bump on the Houston economy, then why does Houston need a federal handout north of $100 billion? Probably because there are distributional issues he has ignored. The people who get rich in Houston are not the ones who have to pay for new homes.

    Also agreed that past a point climate effects get so severe that money becomes less useful in dealing with them.

  96. Ray,

    Probably because there are distributional issues he has ignored. The people who get rich in Houston are not the ones who have to pay for new homes.

    Indeed, and this is – I think – a general problem with the “just get wealthier” narrative. There is no guarantee that those who do get wealthier are those who will suffer the largest climate impacts. There’s also no guarantee that those who got wealthier will feel any obligation to help those who suffer most, even if they are the ones who contributed most (by emitting most) to the resulting climate impacts.

  97. wmconnolley says:

    > no guarantee that those who got wealthier will feel any obligation to help those who suffer most

    Indeed; which is why that isn’t the idea. The idea is that everyone gets wealthier. Over the 50-year timescale, this is true, so that part of the plan is working.

  98. dikranmarsupial says:

    Is Bangladesh becoming wealthy at a sufficient rate to be able to cope with the likely future environmental problems it faces due to anthropogenic climate change (largely not of their own making)?

  99. wmconnolley says:

    > Is Bangladesh

    Likely not; they need to sort their governance out. I am no expert on Bangladesh, though.

    > largely not of their own making

    True, but. (a) You can’t just use that as a blanket answer to everything; and (b) they don’t have “clean hands” (see e.g. http://scienceblogs.com/stoat/2017/05/05/4698/) so in theory get to sue for damages, not cessation.

  100. dikranmarsupial says:

    “Likely not” it isn’t working then (at least not on a global scale)

    “they need to sort their governance out.” given their resources and population with optimal governance would they be wealthy enough to deal with the likely environmental problems by themselves? I’m no expert on Bangladesh either, but being one of the “poorest nations of the world, with a very large population, much of which lives on low-lying river delta prone to flooding, it seems unlikely.

    “True, but. (a) You can’t just use that as a blanket answer to everything;”

    why not? I am pointing out that from a political (rather than purely economic) perspective, there is a matter of natural justice involved. Looking at climate change from a purely scientific or economic perspective is not likely to solve the problem as it has other dimensions (e.g. political) that are a greater barrier to progress (c.f. Hulmes’ book).

    “they don’t have “clean hands” I did say “largely”, the fact that none of us have completely clean hands is why I used that word.

    “in theory get to sue for damages, not cessation.”

    see my comments about option (iii) earlier, that would be an excellent argument (for most it would be incompatible with the golden rule).

  101. wmconnolley says:

    > golden rule

    Ah, but that’s kinda my point. They are showing no particular restraint in emitting GHGs. So they are acting just the way others acted. Everyone is treating everyone else just as they treated themselves: with no restraint (other than some very recent and rather minor restraint, except for a few counter examples. Germany has sunk endless Euro’s into subsidy, but that’s not a model to follow, either).

  102. dikranmarsupial says:

    I note you haven’t responded to my substantive points,

    “Ah, but that’s kinda my point. They are showing no particular restraint in emitting GHGs. “

    (i) citation please (I don’t know either, but a brief web search suggests it may not be true).
    (ii) what choice to they have?

    It is obviously unreasonable to have the same expectations for Bangladesh as for first world nations that have already had the benefit of fossil fuel driven growth and who do have other options.

  103. wmconnolley says:

    > my substantive points

    Oh, sorry. I’d blipped over “given their resources and population with optimal governance would they be wealthy enough to deal with the likely environmental problems by themselves?” The answer is “Yes”. I don’t think it is a very interesting point, though. The difference between real-world and optimal governance is so enormous.

    (i) declined; (ii) I didn’t pretend offer any solutions (other than the obvious “improve govt”).

  104. dikranmarsupial says:

    William Connolley asserted Ah, but that’s kinda my point. They are showing no particular restraint in emitting GHGs. “

    I (reasonably) asked “(i) citation please (I don’t know either, but a brief web search suggests it may not be true [link to UN climate change secretariat data suggesting that Bangladesh had actually reduced emissions rates, when land use change is included]).”

    William responded:

    “(i) declined; “

    making arguments without caring whether they are true is more or less the definition of bullshit. Not being able to retract the claim when your bluff is called isn’t encouraging either. I don’t know what you are hoping to achieve by it, but I had expected rather more from you.

  105. “AIUI everything above the quantum level (and maybe even there) is deterministic, there is no such thing as randomness (only uncertainty).”

    That’s not the way the term is used. It’s almost always used in the context of an ensemble setting. I will give some examples.

    The compelling issue that I have found recently (and it has been heightened in awareness by the recent solar eclipse) is the deterministic impact of the lunar orbit on various geophysical phenomena. First consider that the moon will trigger earthquakes at specific times depending on the exact geographic location of the earthquake. Although long conjectured, this has been verified only within the last year by two different groups doing statistical studies on earthquake timing. This is interesting because it is a mix of determinism and randomness. The randomness is that there are an uncountable number of faults that can trigger at any one time. The determinism is in the fact that the windows in space/time which they will get triggered is highly predetermined. Anything outside that window is greatly diminished in probability of occurrence.

    This is an ensemble setting because the set of earthquakes is the equivalent of an ensemble of particles when modeled by statistical mechanics. The moon is the equivalent of an electric potential that is dragging around a set of charged particles. In more detail, the fault triggering is analogous to the critical threshold voltage Vt in a diode’s operation. When the applied voltage to the diode is increased beyond the threshold, the probability of electrons that are able to cross the barrier increases exponentially.

    Now contrast that with three other geophysics phenomena that may have a connection to lunar forcing. These are ENSO, QBO, and the Chandler wobble. None of these are really ensemble behaviors, because they are each standing wave modes of a single collective state. Consider that there is only one ENSO state that can exist at any one time. Same for QBO and Chandler wobble. In other words, these are singletons while the earthquakes are multiple instances in a large ensemble.

    The outstanding question is whether the cycles of ENSO, QBO, Chandler wobble, and Ocean tides are synchronized by a deterministic external force such as the lunisolar cycle. Except for the last case (obviously), these are still open questions, with a weak consensus that they are each the result of some kind of stochastic resonance, similar to the response to a resonant circuit. This means that the cycle of each of these behaviors is defined by the properties of the medium (ocean, atmosphere, or mantle for ENSO, QBO, Chandler wobble respectively) as opposed to the driving cycle of the sun and moon.

    Now it would be easy to reject the influence of the lunisolar cycle, as all one has to do is show the cycles of ENSO, QBO, and CW are incommensurate with that cycle. The problem is that you can’t, because they ARE synchronized if you analyze the aliasing correctly (astrophysicists know all about aliasing). As Richard Lindzen stated “For oscillations of tidal periods, the nature of the forcing is clear” and “It is unlikely that lunar periods could be produced by anything other than the lunar tidal potential”. That is the nature of determinism and an example of the causative agent that Lindzen spent a career trying to prove with his research on QBO (he never succeeded AFAIK).

    What hypergeometric can bring to the table are tools such as AIC and BIC that can confirm that the causal connections are deterministic and not stochastic. Science never had to use these tools for proving the determinism of ocean tides, but they may be necessary for other phenomena.

  106. “The randomness is that there are an uncountable number of faults that can trigger at any one time.”

    That isn’t randomness, it is a lack of knowledge about the details of all the faults (uncertainty is not the same thing as randomness).I think you would be better off giving your definition of deterministic rather than using examples that use the definition.

    ” It’s almost always used in the context of an ensemble setting.”

    I don’t think that is true. A double pendulum has deterministic behaviour even if you only have one of them.

    “What hypergeometric can bring to the table are tools such as AIC and BIC that can confirm that the causal connections are deterministic and not stochastic”

    I don’t see how AIC and BIC can confirm anything of the sort.

  107. FWIW I view deterministic as meaning the behaviour of the system is fully determined by the initial conditions (and physical laws, obviously). In that case a earthquake is completely deterministic – if you set up the initial conditions (the state of the uncountable number of faults) in exactly the same way, you will get exactly the same sequence of earthquakes.

  108. “* over-parameterisation is not necessarily a problem if you have a prior over the function of a model, rather than over its parameters. We also try to minimise the number of “Twiddle factors” (i.e. things not directly inferred from the data by the algorithm), which is where a lot of the problems with inference lie, and putting a prior over them and marginalising is a sensible approach (IMHO).”

    Interesting to follow the recent discussion on the Hawkmoth effect (in contrast to the Butterfly effect). The idea is that climate models fail in certain cases, but not because of the initial conditions problem but because the structure or function of the model has uncertainty.

    Consider the twiddle factors involved in predicting the path and timing of last month’s solar eclipse. This was arguably all twiddle factors based on prior measurements of the orbits. In other words, the parameters were twiddled until they fit the measurements. Then determinism takes over and that’s where the prediction comes in.

    Maybe a better example is the twiddling that goes into GPS time with respect to leap seconds. Or is this tweaking versus twiddling?

    So what’s the distinction between “a prior over the function of a model” or “over its parameters” ?
    Or are they actually intertwined in many cases? As what happens with sophisticated ocean tidal prediction models: the parameters essentially are the model. The basic model is a sinusoid with an amplitude, frequency, and phase. So the function is the sinusoid and the parameters are the parameters. There is no way to change the function of a sinusoid. They can make it nonlinear, but what happens is that this just turns it into another simusoid with harmonics of the fundamental period. That’s why tidal programs can turn into hundreds of parameters. The set of fundamental sinusoids interact in a combinatorial fashion, thus expanding the parameter state space.

  109. “So what’s the distinction between “a prior over the function of a model” or “over its parameters” ?”

    see the literature on Gaussian Processes for an elegant example (closely related to krigging), where a covariance function effectively expresses a prior over the behaviour of (pairwise evaluations of) the function. You can take a “parameter space” view of such a model (where you have something a bit like a neural network with a prior over the parameters) or a “function space” model where you don’t even consider the parameters and instead the covariance is a prior over the behaviour of the model.

  110. “I don’t think that is true. A double pendulum has deterministic behaviour even if you only have one of them.”

    Even more interesting is an inverted pendulum driven by a cyclic translation of its base:

    “This inverted pendulum was realized by only oscillating its pivot in 58Hz. It can be done without feedback control. You can find the principle explained in some textbooks about non-linear dynamics. This equation of motion is classfied as Mathieu equation. By solving this equation, you can find the stable condition to keep an inverted pendulum upright. See also WikiPedia, which has short description about how it works; http://en.wikipedia.org/wiki/Inverted_pendulum

    The nonlinear Mathieu equation also describes sloshing of water volumes, known since Lord Rayleigh explained Faraday’s wave measurements. What Rayleigh found is that periodic forcings can lead to a period doubling that is self-sustaining (yet obviously metastable). This is the area of ENSO dynamics that has been barely scratched in the research literature — that of annual cycles leading to a metastable period doubling. The bottomline is that not every nonlinear formulation leads to a chaotic, yet deterministic outcome. If it was truly chaotic, it would be impossible to predict, but the periodic forcing may stabilize the response leading to a behavior that is predictable. That’s what I am working on.

  111. “That isn’t randomness, it is a lack of knowledge about the details of all the faults (uncertainty is not the same thing as randomness).”

    Thanks, I got that. Very similar to the Markov math for reliability prediction where one lacks knowledge about the details of the physics and environment leading to failure.

    Google
    https://www.google.com/search?q=markov+modeling+for+reliability

  112. Willard says:

    > The idea is that everyone gets wealthier.

    This isn’t a clear idea at all. It covers cases like one big orange duck getting all the extra money and more while everybody else losing most of their assets. I doubt that this scenario would leave us better than we are right now.

    If honest brokers “normalize” damages to sell that hurricanes are no big deal, they ought to “normalize” wealth in one way or another.

  113. wmconnolley says:

    > > The idea is that everyone gets wealthier.

    > This isn’t a clear idea at all.

    Yes it is.

    > It covers cases like one big orange duck getting all the extra money and more while everybody else losing most of their assets.

    No it doesn’t. I said “everyone”. That was a broad-brush statement; it shouldn’t be taken quite literally, due to Variation. But 95, 99% say.

  114. @dikranmarsupial, @geoenergymath,
    Sure, setting deterministic values for twiddle factors. Trouble is, some of these are peculiar to various optimizers, and Such Beasts tend to inhabit the world of non-linear optimizers. Let’s take R‘s trustOptim for example. Let’s look at prec and expand.threshold.ap:

    prec Precision for how close the norm of the gradient at the solution should be to zero, before the algorithm halts. It is possible that the algorithm will not get that far, so it will also stop when the radius of the trust region is smaller than stop.trust.radius. If the trust region radius collapses, but the norm of the gradient really isn’t close to zero, then something terrible has happened.

    Okay, not fair?

    expand.threshold.ap First criterion to determine if the trust region should expand. If the ratio of the actual and proposed improvements in the objective function is less than this factor, the algorithm will consider expanding the trust region. See expand.threshold.radius. Default is 0.8.

    These are not transparent. How does one take a problem-specific characteristic and find the appropriate inverse value to determine what to set these to? In fact, what most people seem to do is try a bunch of sensible values and see what works. How do you define “works”? And why isn’t that just another kind of stochastic search?

  115. @geoenergymath,

    What hypergeometric can bring to the table are tools such as AIC and BIC that can confirm that the causal connections are deterministic and not stochastic. Science never had to use these tools for proving the determinism of ocean tides, but they may be necessary for other phenomena.

    Yes, it’s possible these things can help to some degree, in some analyses, as friends from the complicated world of quantitative population biology and ecology have discovered over the last 15 years. On the other hand (a) AIC and BIC are far from panaceas (see quote below), and (b) causation is still more complicated, as Sugihara and May and colleagues http://science.sciencemag.org/content/338/6106/496 eloquently and sometimes mind-blowingly point out.

    From the most recent edition (2nd) of Simon Wood’s Generalized Additive Models, page 110, 2017:

    In some quarters AIC has acquired cult status as the one true way to [do] model selection, in a similar way to p-values, somewhat earlier. This is unfortunate nonsense. Like p-values, AIC is useful as a rough and ready quantitative aid to selecting between models, nothing more.

  116. @dikranmarsupial,

    … if you set up the initial conditions (the state of the uncountable number of faults) in exactly the same way, you will get exactly the same sequence of earthquakes.

    But you are assuming you can know and then operationally set up the states of those faults with accuracy sufficient to not have them diverge along completely different paths on either side of those settings. It seems more plausible that you’ll see nonlinear effects on faults than linear ones, so such an assumption seems pretty strong.

  117. izen says:

    @-wmconnelly
    Bangladesh:-” “given their resources and population with optimal governance would they be wealthy enough to deal with the likely environmental problems by themselves?” The answer is “Yes”.”

    Unlikely.
    The environmental problems are global in origin. However optimal, the Bangladesh government would have little influence on global policy required to deal with the problems. If Bangladesh gets ten times as wealthy as it is now, it would be as rich as Portugal. That is not sufficiently wealthy to deal with current environmental problems.
    Is ANY nation wealthy enough to deal with current environmental problems autonomously and independently ? Future wealth and optimal governance would need to be much better at dealing with the problems.
    It seems unlikely given that the problems, and the solutions, are not constrained by the borders of governance or the concentrations of wealth..

  118. Willard says:

    ­> I said “everyone”.

    In ‘one Mercedes for everyone,’ we don’t know how many Mercedes there is, nor do we know if that includes governments and corporations.

    In ‘everyone should get wealthier,’ inequalities ain’t taken into account. It’s quite possible to be wealthier while losing assets. Many Houstonians who just lost their homes could still remain wealthier than their fathers. In the end, the scenario under which everybody gets wealthier first and foremost because industries profit from catastrophes while governments increase their debts sucks.

    Infrastructure problems won’t get solved by proxies that usually become Freedom Fighters talking points.

  119. dikranmarsupial says:

    hypergeometric wrote “Sure, setting deterministic values for twiddle factors. Trouble is, some of these are peculiar to various optimizers, “

    Indeed, there are similar problems for getting MCMC to work reliably without operator intervention.

    “But you are assuming you can know and then operationally set up the states of those faults with accuracy sufficient to not have them diverge along completely different paths on either side of those settings. “

    whether the problem is deterministic doesn’t depend on my state of knowledge or control of the initial conditions, just whether future behaviour is defined by them. The point I was making is that ““Regarding twiddle factors: They are most successful when you know the behavior is deterministic.” is difficult unless you can specify what you mean by “deterministic” (and it doesn’t seem to mean what I think it means). I would view “something that depends on initial conditions, but I don’t have sufficient knowledge of the initial conditions to simulate its exact behaviour” as chaotic (which implies deterministic) rather than random.

  120. dikranmarsupial says:

    hypergeometric I like the quote, I would argue “the one true way to [do] model selection” is not to do model selection, but instead to weight each model according to its plausibility (i.e. posterior probability). But then again, I have written a fair few papers on selecting models, so I am clearly a heretic! ;o)

  121. @dikranmarsupial,

    Regarding:

    Indeed, there are similar problems for getting MCMC to work reliably without operator intervention.

    There are a couple of things that can happen to MCMC. There are similar things that can happen to other stochastic optimizers as well, e.g., Spall’s SPSA.

    First, without knowledge of the (possibly multidimensional) posterior density, it’s possible that it needs to get initialized in many locations. This is one place where having lots of chains running in parallel can help. This can affect any stochastic optimizer. However, I would not consider the list of initializations tried to be a “twiddle factor” because, if another set of initializations — itself picked by some prior — is tried and convergence is achieved, that strengthens the case.

    Second, perhaps there’s poor mixing, and that can be seen in the traces, as well as in diagnostics. Perhaps the proposal function is poorly matched. It’s possible to drop back to an ergodic one, and wait a long time.

    Finally, the “final state” of the MCMC can be used to initialize a more deterministic optimizer on the posterior density, assuming it’s available in closed form. If it wanders off, then that’s evidence that the MCMC wasn’t run long enough, or the density is multimodal and nasty. In that case, exploring the multimodal surface is what’s wanted, or perhaps the likelihood needs to be spruced up.

    In the case of things like SPSA, the “twiddle factors” are pretty much limited to the parameters used when calculating the stochastic gradient, although given that in the canonical case these are sampled from a distribution (for SPSA, U-distributions work well, although you still need to tell it how wide), this isn’t quite as bad as estimating a standard gradient numerically. Still, SPSA, can get stuck in local optima.

    I don’t pretend this is a solved problem. In fact, I’m studying using Levy flight foraging as the basis for exploring a posterior density, perhaps combining it with some kind of empirical likelihood or approximate Bayesian computation framework.

  122. dikranmarsupial says:

    “I don’t pretend this is a solved problem.” good job it isn’t or I’d have to find something else to work on! One of the things I am interested in is seeing how robust/reliable automated MCMC for inference of Gaussian Process models can be made (i.e. so you could just throw some benchmark dataset at it and expect it to work reasonably well), which turns out to be rather more difficult than I had anticipated.

  123. @dikranmarsupial,

    Regarding,

    hypergeometric I like the quote, I would argue “the one true way to [do] model selection” is not to do model selection, but instead to weight each model according to its plausibility (i.e. posterior probability). But then again, I have written a fair few papers on selecting models, so I am clearly a heretic! ;o)

    Sure, Bayesian model averaging and ensemble methods are things, but they have their limits, notably on the same contorted posterior densities which give many of the other methods headaches.

    One of the problems I deal with are ML algorithms which are, under the covers, doing some kind of optimization, and make assumptions about the space of optimization. So when one which seems highly successful on the present sample, even with cross-validation, is tried out-of-sample, they sometimes fall on their face.

    I haven’t published at all. I hope to change that, soon. It’s partly personality: It’s nice to get the successful feedback from solving an engineering problem right off, and, then, when you have that, acclaim seems a poor cousin.

  124. “causation is still more complicated, as Sugihara and May and colleagues http://science.sciencemag.org/content/338/6106/496 eloquently and sometimes mind-blowingly point out.”

    Thanks for that paper. I looked at it and noticed that in one case they were looking at the logistic equation, which is highly unpredictable and is often used to show the Butterfly effect. I am thinking that there are many cases that are slightly nonlinear and likely more predictable.

    The other observation is that they apply “Granger causality”, which is the only statistical test that I see cited by scientists that has its origin in econometrics.

  125. ” Like p-values, AIC is useful as a rough and ready quantitative aid to selecting between models, nothing more.”

    That’s it exactly. These criteria act as tie breakers to choose between models that may all work in terms of smallest error, but the winner according to AIC is the one with the fewest parameters, i.e. the most parsimonious.

    Also, if you consider that a conventional tidal prediction model could have 62 parameters that can be twiddled, the AIC of that could be horrible but it wins because the deterministic physics is probably correct.

  126. There is a distinction between using Monte Carlo in searching the state space for feasible solutions, and using MC to perform a physics simulation. Whenever you do the latter, by definition you are not working a deterministic problem. In the former, the MC is just a tool used to find an answer. In other words, I could use another tool to find a solution. Now, whether you ever will find an answer is the other issue that hypergeometric brought up.

    I also like to differentiate between models that are sensitive to initial conditions and those that are not (i.e. the natural response has long since died out). A good example of the latter is ocean tides. One can say that this is sensitive to temporal boundary conditions, as it’s entirely a forced response and the lunisolar forcing keeps the response synchronized.

  127. John Hartz says:

    Here’s another critique of the Stephens’ opinion piece…

    On Climate, Hurricanes, And Growth by Joseph Majkut, Niskanen Center, Aug 31, 2017

  128. Willard says:

    AT linked to it, JohnH.

  129. John Hartz says:

    Willard: But that was so long ago.::)

  130. dikranmarsupial says:

    geoenergymath wrote “What hypergeometric can bring to the table are tools such as AIC and BIC that can confirm that the causal connections are deterministic and not stochastic.

    hypergeometric quoted Wood “In some quarters AIC has acquired cult status as the one true way to [do] model selection, in a similar way to p-values, somewhat earlier. This is unfortunate nonsense. Like p-values, AIC is useful as a rough and ready quantitative aid to selecting between models, nothing more.”

    geoenergymath responded with “That’s it exactly. These criteria act as tie breakers to choose between models that may all work in terms of smallest error, but the winner according to AIC is the one with the fewest parameters, i.e. the most parsimonious. “

    It would be better if you could just accept that you were wrong rather than this not-so-subtle shifting of the goalposts. There is a huge difference between confirming a causal connection is “deterministic” and deciding to choose between models with similar in-sample performance on the grounds of parsimony (note the idea that model complexity is best measures in terms of the number of parameters is a bit old-fashioned – still O.K. but there are better alternatives)”. The first statement seems to me just the sort of thing that the Wood quote was warning against.

  131. dikranmarsupial says:

    hypergeometric “One of the problems I deal with are ML algorithms which are, under the covers, doing some kind of optimization, and make assumptions about the space of optimization. So when one which seems highly successful on the present sample, even with cross-validation, is tried out-of-sample, they sometimes fall on their face.”

    Yes, this happens very often because the cross-validation error is used to tune the hyper-parameters, which means you then over-fit the cross-validation statistic, so it gives a biased peformance indicator. The more hyper-parameters, the more it can go off the rails. It is not uncommon for out-of-sample data to be from a (subtly) different distribution from the training/cross-validation sample (concept drift), which is another reason this can happen. In principal, models that explicitly capture causal concepts in the problem can help with this, but is rather easier said than done (I helped run a competition on causal and non-causal feature selection, and both had difficulties doing better than a regularised model with no feature selection, but that just means it is a field that need some resources put into it as the basic idea is obviously a very good one).

    “I haven’t published at all. I hope to change that, soon. It’s partly personality: It’s nice to get the successful feedback from solving an engineering problem right off, and, then, when you have that, acclaim seems a poor cousin.”

    Publishing work is not (or at least ought not to be) about acclaim, it is about spreading (hopefully good) practice and ideas that other people can use in solving their scientific/engineering problems, just as you have built your work on that of others.

  132. @dikranmarsupial,

    Publishing work is not (or at least ought not to be) about acclaim, it is about spreading (hopefully good) practice and ideas that other people can use in solving their scientific/engineering problems, just as you have built your work on that of others.

    Apart from the This stuff is okay: It checks out factor, spreading good practice and ideas that other people might use can be done in other ways, such as dropping the article onto arXiv.org. It’s tough to otherwise justify having to put up the page charges.

  133. “… such as dropping the article onto arXiv.org. “

    Many of the AIP journals use arXiv.org as a repository for author submissions. So what you do is write the paper and then upload and then provide a link to the upload when submitting a paper.

    Unfortunately the last time I did that, some volunteer editor at arXiv rejected the upload for some reason. Then when I complained by explaining what I was trying to do, I was told I was banned from uploading to arXiv in the future. Unfortunately that makes it more difficult to submit papers to AIP journals.

  134. dikranmarsupial wrote:

    “It would be better if you could just accept that you were wrong rather than this not-so-subtle shifting of the goalposts. “

    OK, I will admit I was wrong, as this is only a semantic argument which can be interpreted in many different ways. People have been arguing over the philosophy of determinism forever. Maybe you can ask Joe Rogan to do a podcast on it.

    What matters more to me is the context that I brought up my observation in the first place. What I am working on is in confirming that the behavior of ENSO is forced entirely by the deterministic path of the lunar orbit. If confirmed, that would make ENSO deterministic in practice, which would happen to be a significant finding.

  135. dikranmarsupial says:

    hypergeometric wrote “Apart from the This stuff is okay: It checks out factor, spreading good practice and ideas that other people might use can be done in other ways, such as dropping the article onto arXiv.org.

    as long as it has solid peer review (so there is at least a basic sanity check of the contents) that is fine with me, the value of journal publication is only in the extent to which there is a meaningful attempt at quality control. arXiv.org is a jolly good thing IMHO. Unfortunately, the rise of predatory open access journals rather reduces the case for journals, but getting rid of predatory open access journals (by changing the incentives for academics e.g. to be quality- rather than quantity- based).

    “It’s tough to otherwise justify having to put up the page charges.”

    There is no good reason for the page charges. It is perfectly possible to run a top quality academic journal where publication is free for both the author and reader (e.g. JMLR). Most of the actual work of running a journal is done by the editors and reviewers (both usually unpaid), so if you can get someone (or more than one) to act as publication editors, then it is just a bit of administrative work, which apparently amounts to around $10K per annum. In fields where authors can be expected to typeset their manuscripts adequately in LaTeX, it can be done, and the benefits clearly outweigh the added work (provided it is shared out a bit). ISTR there are stats journals that do something similar, but I can’t remember the title.

  136. dikranmarsupial says:

    I wrote “It would be better if you could just accept that you were wrong rather than this not-so-subtle shifting of the goalposts. “

    geoenergymath wrote “OK, I will admit I was wrong, as this is only a semantic argument which can be interpreted in many different ways. “

    No, it can’t. Whether “tools such as AIC and BIC that can confirm that the causal connections are deterministic and not stochastic.” has no such interpretation AIC and BIC can only confirm one thing each, namely whether one model has a higher AIC/BIC than another for a particular calibration sample. ”Like p-values, AIC is useful as a rough and ready quantitative aid to selecting between models, nothing more.” – it doesn’t confirm anything.

    Basically you have just doubled down. It seems that this sort of tedious rhetorical BS is pretty much inescapable on climate blogs. I give up.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s