« Holiday pictures | Main | Random Sales, and the elasticity of supply for the option of doing nothing »


Feed You can follow this conversation by subscribing to the comment feed for this post.

So do you people count as experts or analysts?
Merry Christmas, all; thanks for the posts: keep it up in the new year.

Merry Christmas from Saskatchewan!

You chose blocking BHP’s takeover bid for Potash Corp as the “Worst Idea” of 2010. But you gave short shrift to opponents of the bid. They did not just claim that potash was “strategic.”

They identified several tangible costs of the proposed takeover, such as the prospects of BHP pulling out of Canpotex and writing off Jansen Lake development expenses against profits from existing Potash Corp mines. You may not see those arguments as good reasons to stop the takeover, but you should at least address them.

In any case, Potash Corp shares have consistently traded for more than BHP’s offer price since the takeover was blocked. The implication is that shareholders would have rejected BHP’s offer anyway. So, federal intervention appears to have been less consequential than supporters and critics suggest.

To me, the “Worst Idea” was failing to restructure and increase the royalties that Potash Corp and other companies pay to the provincial government. The people of Saskatchewan should be getting a better price for their potash no matter who operates the mines.

"But you gave short shrift to opponents of the bid. They did not just claim that potash was “strategic.”"

I'm not so sure about that - in fact many just kept repeating the word 'strategic' over and over again. But if you want to criticize me for taking on the more vapid of opponents, I think that's a fair criticism.

"They identified several tangible costs of the proposed takeover, such as the prospects of BHP pulling out of Canpotex and writing off Jansen Lake development expenses against profits from existing Potash Corp mines. You may not see those arguments as good reasons to stop the takeover, but you should at least address them."

Well, I was limited by a 75-word count, but even so, none of those are good reasons at *all*. The Canpotex thing is a complete red herring. Since Sask owns the Potash it can pretty much dictate any outcome it wants by raising royalty rates. The whole Canpotex argument is that if BHP pulls out, they'll lower prices. But the government of Sask has a ton of pricing power through royalty rates.

"In any case, Potash Corp shares have consistently traded for more than BHP’s offer price since the takeover was blocked. The implication is that shareholders would have rejected BHP’s offer anyway."

Agreed. So we have federal government interference for something that was likely not to happen anyway. Plus we sent a message to Canadian and foreign investors that Canada is not a place to do business. This is in a long line of absolutely arbitrary decisions on the part of Mr. Clement - another example would be bisphenol A.

"The people of Saskatchewan should be getting a better price for their potash no matter who operates the mines."

Once you acknowledge this, pretty much any criticism of the Potash deal goes away, because all the 'costs' of the deal centre around a loss of control or a loss of revenue, which clearly doesn't have to occur.

Gilbert is quite confused. Preoccupation with climate change has lessened, not due to any doubts, but to those higher oil prices that will move us away from it, just so long as coal is not allowed to substitute for it. They have done what government could not.

From the state-side perspective:

Best idea: Proposing revenue-neutral pro-growth tax reform (such as we've already seen implemented up North) and showing that the budget could be balanced within a few years using only a combination of mild spending cuts and delaying weapon system acquisitions and modernization by a few years.

Worst idea: Announcing that the goal of QE2 was to lower (nominal) interest rates.

Worst outcome: no real action on climate change.

Best outcome: rising oil prices.

Gilbert can bite me. Bah, humbug!

Yep. Your 'worst outcome' was something I was thinking about as well. If nothing happens in that file next year, it'll move up higher on my list for 2011.

I'm embarrassed to admit it's one I *hadn't* thought of - but should have (or as I said in the intro "now why didn't I choose that?!?") . That might be enough to get me ex-communicated from the Green Party.

I must say I was a bit perplexed with Ms Woolley's worst idea entry. Not with the choice itself so much as the closing comment - I would think that what is unknown is... unknown, including whether it is a good or a bad thing. I'm sure it can explain a lot of irrational risk-related behavior on the part of economic agents, but explain why one or another of these decisions may or may not be *mistaken*, I just don't see how.

So the bottom line is, it's all hunches anyway. My hunch of a bad idea: force-feeding the Montreal subway contract to Bombardier (worst than Potash because Charest lost Kamouraska anyway - duh); the good one is harder to find: I guess building a stronger consensus on the need to boost university tuition in Qc is the best I see.

Jolly holidays all! Keep up the great work!

Yvan - it was a vile task to be given as the odds of saying something that looked stupid were pretty high given, as you say, that it's all hunches anyway.

So my choices were intended to make a point about the nature of policy decision-making. With the case for the penny, what I should have added was "and if we as a nation can't get our acts together to make this no-brainer policy choice, there is something seriously wrong with our political and economics decision-making process."

The oil sands tailing ponds are a case of a known unknown: we know that they will have some environmental and health impacts, we just don't know what they are.

They are also unknown unknowns - things that we don't know and we don't know that we don't know them.

E.g.some people believe that AIDS was spread throughout Africa, in part, by the re-use of needles in mass immunization programs.

With immunization programs, there were some things that people knew that they didn't know, e.g., exactly how long the protection against disease would last. Those didn't really cause much by way of problems.

But that they could be spreading AIDS was something that the people giving vaccines didn't know they didn't know.

And it's undone the good of the original vaccination program.

Best idea: MP Roy Cannon's push to amend the Importation of Intoxicating Liquors Act to provide for a personal exemption from the IILA for anyone who wants to be an interprovincial buyer of wine for personal use.

Worst: Clearly the voluntary long-form census idea.

I think you mean MP “Ron Cannan.”

Frances - My point was that unknown unknowns are no more likely to be unknown bads than to be unknown goods. But even when known unknowns are known to be bad, i.e. we just don't know by how much, they're still no help for policy-making: if we can't measure the cost of a decision, we can't measure how much of a benefit is needed to make it a rational one, right?

Oh, and funny you call the penny thing a *no-brainer*. In the usual sense, it sure is, but the fact this has been the case for so long is a pretty darn good example of a typical *brainy* thing, in the sense of our brains making us do things that are at odds with most accounts of rationality. Sort of a natural experiment in measuring the dis-utility of change, I suppose.

Yvan "My point was that unknown unknowns are no more likely to be unknown bads than to be unknown goods."

Economists generally believe that there are very strong forces pushing economies and societies towards efficient outcomes. That doesn't mean that we live in the best of all possible words - sometimes individuals acting in their own self interest do collectively destructive things like polluting the environment.

But the University of Chicago principle - that there are few $20 bills lying on the sidewalk - is not a bad approximation of real life.

So the chance of unknown unknowns being really wonderful good outcomes is low - that would be like a stack of $100 bills on the sidewalk. Doesn't happen.

Put another way: positive externalities are internalized. Negative externalities are not.

Even if your home is on top of a diamond mine it might not benefit you at all financially - someone else probably owns the mineral rights to your property. But if your home has nasties in the walls (asbestos, lead, formaldehyde, radon, knob and tube wiring, etc etc etc) - that's your problem.

So the chance of an unknown unknown being a positive externality showering multiple benefits on you and me is low. The chance of it being a negative externality are much higher.

Frances, sorry to impose like this - maybe we just do not define the word *unknown* the same way. 20$ bills are rare because they are unlikely to be unknown by everyone. The point you make about externality is crystal clear, but unknowns in the sense of true uncertainties (as opposed to computable risks) are a whole different animal, unless I'm mistaken, and this is what I'm talking about. Or maybe we'll just disagree on this one.

Yvan - perhaps we will have to just diagree.

Whether unknown unknowns are likely to be good or bad depends a lot on where one is standing.

Think back through your life - how many times have you ever been hungry because you couldn't afford/obtain food? Other than wilderness canoe trips, I can't remember many. And I don't think that's unusual for a Canadian in my generation.

How many people at any other period in human existence - or in any other place on this planet - have so rarely experienced hunger? Cold? Illness?

So from where I'm standing, it seems that the prospect of things getting much worse for future generations is very real, and the prospects of things being better for my children than they have been for me - I'd love to see that happen, but I think big improvements are pretty unlikely.

For people 500 years ago, the calculations were probably very different - and that's why they did things like try to sail across oceans - the unknown unknowns weren't likely to be much worse than what they left behind.

Deprecating the F-35 purchase is easy but misguided. We are going to buy that fighter, period. That outcome was determined when the US took the decision that the F-35 would the the ONLY model it would purchase for its future needs and that Lockheed would be the only fighter manufacturer left. When Britain became a co-developer, the fate of future Canadian procurement was sealed.

There is no alternative. The F-18E is a paper tiger. It's a technological dead end. How exactly are we supposed to get a "competitive quote" when the US and UK can't manage that?

Best Idea: Pharmacare. Would somebody please tell me why having a business/employer act like a "nanny" and provide prescription drug coverage is somehow economically efficient? The business of business is to make money, not act as a welfare agent for its employees. Why can't we admit that leaving drugs out was a mistake and include them like the NhS does in the UK? Where do you think the idea for Medicare came from in the first place?

Why can't we admit that people enter and leave employment frequently, but their medical needs remain, regardless of the labour market? Why can't business just pay a bit more in income taxes and be rid of the welfare burden?

"Would somebody please tell me why having a business/employer act like a "nanny" and provide prescription drug coverage is somehow economically efficient?"

It's really not. It's pretty good example of the distortionary nature of bad tax policy. As an employer, if I give an employee an extra $2000 in cash or a $2000 health policy, it doesn't make too much difference to the bottom line since both are expensed. (That's not entirely true because I might also have to pay employer side EI and CPP on the $2000, depending on whether or not they've crossed the cutoff).

On the employee's side it's a different story, since they have to pay income tax on the cash (and possibly EI and CPP) but NOT on the health plan. So for an employee in Ontario earning $50K/yr the comparison isn't between $2000 health plan and $2000 cash - it's between a $2000 health plan and $1377 in cash after taxes.

The main reason companies offer health plans instead of raising salaries is to take advantage of a distortionary tax system. I think we'd see a lot of these inefficiencies go away if we cleaned up the tax code.

Mike, I must admit that I couldn't figure out the question! Best and worst what? Economic ideas? Economic events? Consequences of past economic ideas put in to practice?

I was very surprised that nobody nominated the Euro as the worst something of the past year. So far the Euroskeptics' arguments of twenty years ago seem to have proved prescient.

As for the best, some of you quasi-moneterist macro types might have chosen the idea of Olivier Blanchard and the IMF to raise inflation targets on a permanent basis so that monetary tools have a better chance of counteracting deep recessions like the current one.

"Mike, I must admit that I couldn't figure out the question!"

Heh, I don't think you're at all alone. Reading through the responses I think the question got interpreted 14 different ways.

Mike agree with me!

I was trying to show why employer health plans are inefficient without mounting my "Lefty Soapbox". Further to his post, in a crunch the first things businesses will try to trim are health plans and retirement provisions. If you want to see what "declining real income" looks like, include both health plans and retirement provision (especially DB pensions, or lack thereof) in the total compensation picture. As I have insurance training, I know of countless way that you can try to "shave the ice cube" by trimming benefits and I have seen plenty of examples of this.

Quebec eliminated the tax exemption for Health Plans in its provincial income tax system in 1997. They moved to the standard regime for insurance policies, that is to tax either the premiums or the payout, but not both. They also mandated universal prescription drug coverage through employer plans and the RAMQ. Employers have to offer a plan and employees must register for that first. The RAMQ is second-payer.

The only advantage to this system over a pure public option is it keeps the group health insurance brokers in business.

Still, once we admit that employer-based health insurance is economically inefficient, the choice is between no system at all or an individually-based insurance system (still inefficient) and a universal plan, which is socially and economically optimal.

Sorry Gregory, but Blanchard's proposal deserves the obscurity into which it has fallen. Even Blanchard has backed away from it.

Determinant: Can't really argue with any of that. Except perhaps: "Further to his post, in a crunch the first things businesses will try to trim are health plans and retirement provisions. " I can think of a few things that cut first. Like excess employees, for one. But if cash wages are stickier than non-cash wages, then this makes sense.

I'm the wrong person to be discussing drug plans as I tend to believe our society is overmedicated as is. As this doesn't come from any formal training, it's probably not appropriate for me to discuss on this forum. But if you change 'drug plan' to 'preventative health plan' I'd be more inclined to agree with you. :)

In other words, it'd be more optimal to have our public policy prevent people from getting Type II diabetes than it is figure out the most pill delivery mechanism.

Last point before I shut up about this [I promise]:

If I give my employees a drug plan, it's not considered a taxable benefit. Furthermore, prescription drugs are exempt from the HST.

If I give my employees a gym membership, it's considered income and taxed. Furthermore, gym memberships are assessed a 13% HST in Ontario.

You're going to have a very, very hard time convincing me that we shouldn't fix the tax system first when it comes to health issues.

That's what I was getting at, that cash pay is stickier than non-cash pay.

Re: overmedicated.

Depends on who you talk to. I am a Type I Diabetic so that tends to colour my perspective of drug coverage and other things. As my disease doesn't go away when my work situation changes, I am not wedded to the idea of employer-based drug coverage. I do look favourably at universal-cover systems like the NHS in Britain.

There are different kinds of drugs. There are long-term use drugs like heart medication that tend to be inexpensive, expensive long-term use drugs like insulin (my dependence on it will not go away unless my prayers are answered), and occasional-use drugs like antibiotics.

Not all drugs are created equal in scope or need.

However by all means go for prevention. BTW Type I Diabetes is a far different creature under the hood than Type II Diabetes, the far more common kind often associated with obesity. In my case no prevention was possible. But your point is really a policy, er, political one.

In Canada Group Health Plans generally mean three things: Drugs, Dental and Disability. I want to move drugs into the public domain and we can debate dental. There is some basic need for care but there is also much that can be done individually or isn't strictly medically necessary.

With Disability you are actually insuring an income that is impacted by disability. No group benefits plan in Canada is seen without Disability Income Insurance. That's just the way it is. The real problem is that group Disability benefits are written for the benefit of the employer. They have a restrictive definition of disability ("Any-Occ") that most people don't understand. There are better and more complete options available on the private market which better fit people's expectations, but most people don't know or care about the problem.

I have a good book on how to get a complete DI "Program" but like many things most people don't know there is even a need for it.

BTW I don't sell insurance anymore so I'm not plugging myself.

"Depends on who you talk to. I am a Type I Diabetic so that tends to colour my perspective of drug coverage and other things. As my disease doesn't go away when my work situation changes, I am not wedded to the idea of employer-based drug coverage. I do look favourably at universal-cover systems like the NHS in Britain."

That makes sense. And I'm not wedded to the idea of employer-based drug coverage either. It doesn't make any economic sense.

My sympathies RE: Type I diabetes. My wife does health care policy researcher and she's done a fair bit of research in that area, so I've picked up a fair bit via osmosis. (Incidentally I have a hunch she'd more be inclined to agree with you than with me). But yeah, clearly not all (or even the majority of) medication goes to self-induced conditions.

Not surprised to learn those benefits are written for the benefit of the employer, since the employer is ultimately the one who chooses the plan.

See, that point is lost on most people. Due to quirks of tax law, DI policies whose premiums are paid for with after-tax dollars are tax-free on payout. Therefore most pay stubs in Canada will have an after-tax DI premium deduction. For 3-5 year disability this is OK, but after that it gets hazier. You can get a tax-deductible DI plan that has a taxable payout. This can preserve retirement income. But most plans aren't structured that way. Most people don't recognize what actually goes on in a group benefits plan.

Most people if purchasing a personal DI plan would want what's called a "Regular-Occ" disability definition. It meets most people's expectations and preconceived notions.

Again, lack of financial knowledge on the part of the majority to actually purchase what they want and need.

Frances - I hope you don't mind if I pursue this just a tad further. I think your latest argument is stronger than the previous one, yet I still don't think it's enough to transform unknowns into policy-relevant parameters. I read your last point as sort of a generalization on the decreasing returns story, and I certainly won't challenge your belief as to where we are on that curve, since for one, I do share that belief somewhat (having the same type of pessimist-biased brain as everyone else), and also because beliefs - especially when we can't verify them - are all we have here.

Yet, even granting we'd be pretty close to the top of the marginal product curve of human activity, how can we include the unknown into our assessments of the *relative* costs and benefits of different decisions? Why should the unforeseeable catastrophe come from pursuing our most polluting activities as opposed to stopping them? What if the alternative course we'd pick turned out worse than the current course of global warming? How can the best science be of any help here, i.e. *beyond* what it can tell us - surely, the unknown can not be something that science can evaluate, or do we have a different concept of science as well?

I'm sorry if it looks like I'm attempting to justify some sort of *anything goes* stance - I'm not. I think our hunches do have a crucial role to play in public policy - but I also think we'd have a better public discussion about these issues if we agreed that no one has a privileged access to truth beyond what science can in fact tell us.


I think your position is almost absurd.  Take any highly organized system; do something random to it. The odds that you will do harm vastly outweigh the odds that you will do good.  Life didn't *evolve* because of random mutation.  It evolved because of natural selection. For every gazillion mutations that got destroyed in an evolutionary instant, one or two provided the bearer with some adaptive advantage.  Same goes for random pollutants.  Consider the odds of pouring some random chemical into a lake or into the human body and it turning out to be a really great thing. Random significant change is *bad* almost by definition.

The debate over global warming is already way too skewed in favour of the polluters.  Somehow the burden of proof on human-caused harm got shifted onto the environmentalists instead of the polluters.  Given any sane Bayesian prior of possible effects, that is totally insane.

That's my vote for economic stupidity of the decade.

K - I don't think that the nature of evolutionary processes provides any help with criticizing my point either. Take any 2 of those gazillion mutations, and you're still stuck with an uninformed bayesian prior of a 50-50 chance that either will be more adaptive than the other. That's really all my position entails. Where is the absurdity exactly?

Remember I'm only arguing about how the unknown is unhelpful as a policy guide - I'm not at all suggesting that science, i.e. what is known, is at any loss to support good arguments for a strong environmental agenda.

On what was apparently deemed to be the dumbest "economic" decision of 2010:

The purchase of the F35 fighter. Any future deployment of our troops will be against insurgencies, so buying these fighters makes no sense particularly since we do not have a coherent longer-term defense policy. We need patrol craft for our coastlines not toys for the boys.

That doesn't sound very much like an economic justification/refutation. It sounds very much like someone who doesn't think about defence matters weighing in on one. The fact of this being a stupid economic decision is apparently supported by the "fact" that air forces are irrelevant, but I don't remember that having been established as a fact anywhere.

Wouldn't it be something if public intellectuals stuck to their own areas of expertise?

Yvan:  But you said: "Why should the unforeseeable catastrophe come from pursuing our most polluting activities as opposed to stopping them?"  So we are not talking about two different random mutations.  We are talking about either mutating (dumping CO2 into the atmosphere) or not mutating (not dumping it).  

Maybe drinking random chemicals is a better analogy.  Your choice is "drink it or don't drink it."  Not at all equivalent to "drink this one or drink that one".

I agree that if a human had decided to drink a chemical, then who are we to say that the unknown bad consequences are worse than the unknown good ones. The prior assumption should be that that person has already weighed all the outcomes.

But pollution is more like someone deciding to give other people random chemicals, with *no* consideration for the health impacts.  The prior, in this case, is inordinately *bad*.

Geoff: That one struck me as odd as well. To be fair, David Bond may have expertise in that area - I don't know.

The F-35 is designed to provide Close Air Support, that is bombing/strafing random enemies in support of our own troops. Plus we'll use it as an interceptor, and "stealth" is being designed into every military aircraft these days, regardless of what said aircraft will be used for.

As with most military procurement, there is two parts politics for every one part of informed debate.

K - it's just a matter of semantics, then. We're not really disagreeing. Much of our misunderstanding has to do with how rigorously we use the word *unknown*, and I've also used *unforeseeable* in the same strict sense. It has nothing to do with either what science tells us will be the consequences of something, or even what it tells are likely consequences. On the contrary, it simply refers to what we have no clue about.

In your example with chemicals, you choose *randomly* out of set of chemicals among which you *know* that the ones you can drink safely are in a minority. But you're using what you know to make this argument, not what you don't. Further, if you know that one of them is essential to remain alive, drinking one at random is surely more rational than dying of thirst. But I digress.

The real point, to get back to *polluting activities*, is that using that very phrase implies only what we know about them, i.e. that they have some known bad consequences; it implies nothing about what we don't know about them. And what we don't know about drinking water may be just as bad as what we don't know about drinking sulfuric acid. What's good or bad about drinking these things is something we do know (insofar as we can act on it). Same with pollution. And the logical consequence is that among all unknown consequences (including unforeseeable catastrophes), we have no a priori reason to blame any on either polluting activities or non-polluting ones. It's not a matter of science, but one of logic and definitions.


The point is not that I chose a particular set of chemicals. The point is that we chose a system (economy, ecology, human body, etc) in a highly organized and optimized state. I.e. there is nothing random about it. If you change it in a random way (pick a random unknown chemical), it is, with *extremely* high probability, going to be less organized and less optimized than it was before. So by "polluting" I *don't* mean "anything known to have bad consequences" - I mean *anything*. Do *anything* of consequence and it's *inordinately* likely to be harmful. Walk into a tidy room. Move an object to a *random* place. It is overwhelmingly likely to be in a place that's not appropriate for it. That's polluting. Create white noise, or say random nonsense when someone is having a coherent conversation. That's also polluting. Being orderly, organized and optimized means being disrupted by almost anything.

Take a chemical chosen randomly from all the chemicals in the universe that are unknown to humans. I challenge you to drink it until such a point that you notice some observable impact. Is it your contention that your expected positive utility balances your expected negative utility? Seriously???

Your original point was about tailing ponds. Lets assume that scientists have made some list of "effects of tailing ponds". Despite my knowing nothing about tailing ponds, I'm going to venture a guess that there is nothing positive on that list. How do know that? Because I know that the ecosystem that was in place before was adapted by natural selection to an environment *without* tailing ponds, and tailing ponds appear to be a significant deviation from that previous order. I don't need to know whether "tailing ponds" contain petrochemicals, bubble gum or Unobtainium. I makes zero difference to my prediction. All I need to know is that the previous state was in some sense optimal and not random. The subsequent state is more random and therefore less optimal. So when it comes to making predictions about all the unknown unknowns, i.e. all the things the scientists didn't put on that list, because, well, they didn't know about them, I'm going to guess that those things will be bad too.

Your point about water is exactly wrong. Whatever we don't know about water, we are adapted to drink it, and therefore overwhelmingly likely to tolerate it. Unlike sulphuric acid. Same with pollution.

As far as "non-polluting" activities (things that don't change the environment), yes, we do in fact have excellent reasons, a priori, not to blame those. The polluting ones, not so much. It is exactly about science and not about logic or definitions. It is about the second law of thermodynamics and its affect on organized systems: life is a struggle to maintain highly ordered systems against increasing randomness or "entropy" which wears us down and ultimately destroys us. And that has everything to do with science.

K - optimality is unobservable. It can be useful methodological tool, but that's it. If you take it as a description of the actual world, however, that belief is of the exact same nature as the belief in God. And this is precisely what I think should have no place in this debate. The rest of your argument is still all based on using the word *unknown* in a different sense than *about which we know nothing*. Again, it's not what is unknown about your random chemicals that makes them a risky proposition, it's their being known as part of the class of liquid chemicals, only a small part of which may be absorbed safely.

By the way, I'd say our disagreement proves one of 2 things: either there is something sub-optimal here, as you and I can't both be right on this, or else our views and everything else are all causally determined to be what they are and there are no degree of freedom to do anything else than what we do, agreeing or not, polluting or not, random or not. I don't think you can really have this cake and eat it too.

The funny thing is that you're saying my position is absurd, yet you start your case for changing something (i.e. having a stronger environmental policy for example) with the assumption that everything is wonderful and that changing anything can not be a good thing. How can these 2 ideas be at all compatible?

Yvan:  You have already accepted the idea that some states are more optimal than others if you allow discussion of utility.  But, that is why I said "in some sense optimal".  I.e. the duck being alive is "in some sense" more optimal than the duck being dead.  And that state of being alive depends crucially on a huge number of variables with very particular values.  And if any variable changes randomly, that is far more likely to be less optimal (dead duck) than more optimal.  But if you don't like the word optimal, just substitute "highly organized", i.e. far away from maximum entropy.  The earth is full of systems like that.

I never said that "changing anything can not be a good thing."  I said that changing a random thing in a highly organized system is (a lot) more likely to be bad than good. How can you ignore this fact? Keep changing random things and eventually you will be at maximum entropy, i.e. your system will be a perfectly uniform mess, devoid of any structure, at which point, we will in fact be indifferent to random changes.  But not before then.

K - I'm not ignoring any fact. I am precisely not disputing facts. I'm arguing about what science *can not* say anything about. And I find it fascinating how what seems to me rather trivial about it is taking us into a huge argument.

Actually, I don't even think that the concepts of organization and random disturbance are relevant at all to my point. You arbitrarily pose an organization and an external source of disturbance, and say a disturbance is a disturbance. Well, granted. Now what if the source is a part of the system, and it uses its degrees of freedom to make decisions based on the information it gets from within the system. All I'm saying is that the information that is not available is of no use to that decision-maker. Am I missing something up to that point?

Then, it seems to be a matter of sheer logic to deduce from that, us being such decision-makers, that the unknown consequences of any decision are just that, unknown. And we're always discussing about decision A as opposed to decision B, we're not talking about *disturbing* decisions taken in vacuum or as opposed to no decision at all, are we? I have no clue how that could have any concrete meaning.

Yvan: I agree it's pretty weird. We both seem to find our perspectives trivially obvious, which makes the debate more difficult. I wouldn't normally persist, but I feel like this has important consequences for how we weigh the environmental impact of economic decisions. So (sigh) here goes...

"Am I missing something up to that point?"

Maybe you are missing that there is not just one single fully optimizing agent making and bearing the costs of rational decisions. Some agents make the decisions; others bear the costs. And when government scientists assess the expected damage, their reports only assess the costs of the enumerable set of things of which they are aware. I.e. there's a list of concrete issues with associated expected costs. And stuff not on that list gets zero consideration. The way the system works, the party that is being harmed can't just say "pay me for all the things that could go wrong but that we haven't thought of". If there were a single rational agent, it would take into account the fact that the unknown unknowns (i.e. stuff not on the list) are more likely to be harmful than good, and, expecting to bear those costs as well, would take them into account in their decision making. In that circumstance it could be argued that the residual estimation error is random. But in the real world, there is no one assessing the expected costs of the unknown unknowns. The expected costs assessed by the government scientists therefore represents a lower bound on the real expected costs.

I (and I think Frances) define the "unknown unknowns" as everything that cannot appear on a list because nobody ever thought of it. But if I'm right, then you are just defining "unknown unknowns" as estimation error on our best estimate of *all* effects (both on and off the list). But if that's the case, I would have just called it random estimation error. I also think it's not a very realistic way to think about pollution because nobody performs such an estimate unless the impact of the decision is only on themselves.

I hope that gets us closer :-) If not, at least let me know how you define "unknown unknowns." Also (assuming you disagree with my definition) do you agree that *if* you accepted my definition of "unknown unknowns" then those things have expected negative utility?

K - first, let me welcome this opening on your part. But getting to speak the same language does not appear any easier. Already I was a bit taken aback by the move from the uninformed to the random, and now I'm just as unsure about the consequences of moving from the unknown to the *never-thought-of*, but I'll still give it a try.

Clearly, something that's never been thought of would belong to the category of what is not known. Technically, knowledge is generally defined as the set of *true beliefs* that we entertain at some point in time. And this is my starting point (I'm assuming that we have a working, applicable definition of what is true, obviously, otherwise we'd have a whole other ballgame on our hands). Hence, what is not known includes both what we believe falsely and what we have formed no belief about. The *never-thought-of* would thus belong to the latter. The *don't give a damn anyway* could also form a similar category. At any rate, I see nothing in what I believe about what is unknown that I would not believe about what has *never been thought of*, since I see the latter as a subset of the former. And really all I believe about it, is that it is unknown, uninformative, and useless (well, as long as it is unknown, obviously).

Also, like I said in an earlier response to Frances, I see no real point in distinguishing known from unknown unknowns, as their being unknown is sufficient - in my view - in declaring them useless for rational decision-making. My main contention is that choices are always between alternatives, and that what is unknown is just as unknown for one side of the alternative as it is for the other. And whether the expected value of what is unknown is negative or not is simply irrelevant, as what is unknown, hence as much of a negative as you may want to believe, it is just as unknown on both sides of the alternative, so that relatively speaking, it can not weigh any more on one side than on the other.

What I hear you saying, K, is that what you know about what is good or bad should be enough of an indication that what you don't know "behaves" the same way as or at least in some positive relation to what you do know, and this is what I disagree with. I don't know where you get that knowledge about what is unknown - it just sounds like a contradiction in my ears.

In other words, when comparing advantages and costs of any decision, I hear you saying that all cost assessments should be multiplied by some totally arbitrary factor, because what is unknown is more likely to be negative than positive, so that all negative units should count for more than positive ones. Beyond the arbitrariness of such a factor, which is a enough of a problem in itself, I just don't see how the underlying concept can be supported by either your previous examples, or your reference to systems theory, or by any factual aspect of what we know about the real world.

What I'm saying basically is that what is unknown is uncorrelated to what is known. If it was, then it would mean we'd... know something about it.

Sorry for being so obtuse - I know I'm not conceding much. But to me what's at stake is no more and no less than the meaning of the word *unknown*. It can't rely on anything that we know, with the exception of the english language.

The worst was cutting Stats Can data collection in some areas including the long form census. The best was the increase of payroll taxes but they don't take effect until the new year.


I think we are no longer in the same conversation.  I feel like I'm talking about the practical limits of scientific analysis in the economics of pollution. I think you are performing a post-modern deconstruction of the framing of epistemology in the English language :-). I'm sure we are both making excellent arguments, but I doubt that they have much to do with each other.  I did enjoy it for awhile (thanks!), but I am losing faith that we have any hope of ever converging.  Till next time!


Alright.  Maybe I'm a glutton for punishment, but I do feel like we've fought a valiant fight.  And maybe, just maybe, I've figured out a new move, though I suspect you will deftly parry as always...

A scientist writes a report outlining the risks associated with tailing ponds.  It contains an enumerated list of things that will or may cause harm.  One of those things is that ducks may land in it, but the scientist doesn't know how many ducks will land in it.  So she models the number of dead ducks as a random variable with a particular distribution derived from the known migration pattern of ducks.  And she calculates the expected number of dead ducks from that distribution.  And if she is really helpful, she even calculates the expected monetary utility of dead ducks, so that she can coherently aggregate the cost with the other risks.  This is what I would call a "known unknown".  It is "known" because we are aware of the possibility of ducks landing in the lake.  It is "unknown" because it is a *random* variable.  I.e. we have no way to know how many ducks will actually die in a tailing pond.  But we are able to describe the distribution of that variable.  So by "known unknowns" I just mean random variables in our model.

Years after the report is done, humans are infected by a strange bacterium found to have been transmitted from ducks which survived limited exposure to tailing ponds.  The bacteria which previously fed off tar deep below the surface of the tar sands, incredibly finds itself happily at home in the human body.  This is an "unknown unknown", the first "unknown" meaning that the sentient agents whose utility is being optimized were not aware that this could happen; the second "unknown" meaning that it's a random variable, i.e. even if we have a good idea of the distribution of possible outcomes of this variable (now that we know it exists), we don't know what will actually occur.  Unknown unknowns are very different from known unknowns because there is generally no constructive way to quantify them.  But the fact that we have no solid theoretical way to model them doesn't make them go away in the real world.  In the real world shit happens. [It is my contention that these "unknown unknowns" are generally bad, but that is a corollary that is not the point of this particular comment.]

At the risk of diverting your attention from my concrete example, I'm going to make a more general point: My objection to your line of reasoning is the same as my objection to all of the rational choice literature.  Real rational agents are required to make simplified models of the real world.  The process of constructing those models is Bayesian.  I.e. you start with a vague prior of all the possible correct models of the world, and then through observation and induction you refine your prior.  But there is no unique choice of prior, because there is no unique way to define the set of all possible coherent models of the world.  And therefore, in the real world, there are no unique rational expectations.  Just different people each with their own subjective, yet rational, expectations.  To bring it back to our conversation, "unknown unknown" just means "not in the world model of the utility optimizing agent" in the very specific sense of the bacterium discussed above.

Stephen: oh well. I think I'm on firmer ground nominating the Euro as a bad idea, and here's some backup hot off the screen.

go K! You've got it all right so far.

Yvan, your arguments sound awfully tautological.

Maybe a little help: the risk with the "unknowns" is that we don't know what the outcome will be, but have a general idea of what (even in the extreme case) is possible; the risk with the "unknown unknowns" is that we don't even know what the possible outcomes look like (we've never seen them before!!).

These two risks are not the same. Try to guess which one is worse!

Thanks for the encouragement, Sina!  Much needed as when running a marathon.  

I agree with your comment, and might rephrase it like this:  For the known unknowns we have a good idea of the support of our distribution (e.g. the set of integer numbers of dead ducks) and even some idea of the measure of that set.  For the unknown unknowns we don't even know the support.

As for which kinds of risk is worse, I think it depends how well we know the system.  But in many environmental cases (e.g. CO2), you might be right.

There is, it turns out, a really fun wikipedia article on unknown unknowns. Apparently this contribution to epistemology, was not in fact due to Rumsfeld. The term has been in use in the US military at least since the '80s. But the article also mentions another important category of truth: the unknown known. These are things that deep down inside we know to be true but which we repress from our conscious decision making (the potential for the events at Abu Graib was the military example). In common terminology, I guess we call it the elephant in the room, an often it's probably even more dangerous than the other unknowns (eg "the rating agencies are beyond reproach").

K - I don't really know why you say we're *no longer* in the same conversation. We certainly both have been saying the same thing all along. And I just can't see how I could change what I'm saying, because the very essence of it holds entirely in the following proposition: the word *unknown* refers to what is not known. I feel like you'd want me to say: oh wait a minute, no, you're right - there are some known things that are also unknown. Like blue things that are also non-blue. And you're saying I'm doing postmodern deconstruction? Is this a variant of Aristotelian logic that I was not aware of?

Sorry to be so facetious, but I'm sure you know as well as I do that it's getting real close to when we must call a couple more beers and switch to whether the Canadiens will finally get the friggin Cup this year. Yet I will honor your efforts by restating my case once again, in light of your new re-statement of yours.

First though, let me welcome as warmly as I can your new ally: Sina, your assessment of my argument is perfect - it *is* awfully tautological. I couldn't have said it better. This is precisely what makes this conversation so strange: a tautology is always true! Why the heck am I defending it, and why is K attacking it? Your guess is as good as mine - dunno. Oh well, I'm sure we could make all sorts of anthropological hypotheses about it, but none of them would be verifiable, would they.

So back to work. For one thing, I remain reluctant to use the *unknown unknown* versus *known unknown* terminology, because although it may have literary merit, there is a simpler, less confusing way to express that dichotomy that goes back to ole Frank Knight's distinction between the risky and the uncertain, the first being amenable to some probability distribution that you can use in decision-making, the other not being so. Now, we can reword this by saying that some phenomena, now or in the future, can be described as a mix of some things that are known (the evidence from which a probability distribution can be derived) with others that are not (how it will in fact turn out), and yet other phenomena include no known elements at all. Yet in all this last sentence, the phrase *not known* has always referred to the same type of quality, and about which nothing can be said apart from its being unknown, since it is not known what else it could be.

To me, the story should end there. But you seem to maintain that you actually know something about what is unknown. And not only are you telling me that its overall expected value is negative, but you're telling me that this value is significantly correlated with the expected value of what is known. I've separated this problem in its 2 parts, and suggested I was ready to grant the first one for the benefit of the conversation, also because I respect what I see as expressing a legitimate desire to rationalize one's pessimism, and lastly because it doesn't really affect my central point, which is that it changes nothing in terms of decision-making. The crucial part of the problem, in my perspective, is the correlation bit.

Now the way to counter your latest attempt is thus pretty clear, I think. You suggest that maintaining or expanding these ponds could generate some infection that we'd become aware of years after. I could counter your assessment of the expected value of what's unknown by suggesting that this infection was also discovered to have allowed humans to build new antibodies that finally explained the mysterious and dramatic recent drop of more than half of cancer incidence.

But don't let me be so utterly and disgracefully optimistic. Let me rather suggest a different alternative story. This way I can share your general pessimism, yet still defend my main contention. I even offer 2 variants for the price of one. Here goes: your story about this bacteria got Albertans so darn outraged that they elected the Greens and got rid of all this dirty stuff. Years later, it was discovered that, variant A, some chemical process in the ponds was actually keeping the mean bacteria deep under the ground, which explains why people are infected by it now (and wouldn't have been if the exploitation had gone on), or variant B, as humans are also part of the story, some bitter laid-off workers then colluded with some crazy terrorists, helping them to find the way to poison Toronto's water supply that they had been looking for for so long. (I don't if you got this one in English: "l'enfer est pavé de bonnes intentions". Inferno is paved with good intentions? Ah, those darn unintended consequences...)

Hey, how do you like them apples? Are we calling these beers or what!

Yvan:  Yes, we definitely seem to be closer to the same language now (i.e. Go Habs, go!).  And I take back the post-modern accusation.  It was, in fact, an underhanded slur. While I'm frustrated that you wont accept my epistemology (yes, you were supposed to say that I was right and you were wrong...), I'm happy to move on to concrete examples.

But first, I don't understand why you say that I'm assuming correlation between the knowns (risks) and the unknowns (uncertainties).  I'm not aware, at least, that I was.  I thought I was assuming they were independent.

Taking the example of the bacterium, I was hoping you would say that it might cure cancer. But unless it was sent by God to cure cancer, it's just going to go about its selfish business of feeding and reproducing.  Ignoring the obvious negative impact that it is competing with the other cells in the body for resources, it may either 1) not interfere with any processes in the body or 2) interfere. If we assume that most of the processes in the body have positive utility for the human, it's simply vastly more likely that it will cause harm than cure cancer or any other good. The odds of good versus harm are comparably dismal as the odds of a random mutation being good versus harmful.  This is a fairly straightforward result of the system being in a highly particular and optimized state *from the perspective of the owner*.  At this point, I think you would object by saying that I am using some information that I *know* about the system.  Yes, it's true.  I *know* that the system is very particular.  But I'm assuming nothing about the "uncertain" thing that might happen to me except that it might do *something* to me.  And *something*, which includes anything that hasn't been part of my environment before and for which I'm not already adapted, is likely to be bad.

As far as your other stories go, yup, I agree both good and bad things can happen totally unexpectedly.  But you have made no case regarding the relative probability.  I'm saying that random unexpected (uncertain in the the sense of Knight) events, even if they are equally likely to cause wealth increase or decrease, will cause a rearrangement of wealth which, given that the original state was Pareto optimal relative to the utilities and world models of the agents, is overwhelmingly likely to be less optimal.  I've been trying to construct a mathematical proof that this is true. But, I'm having a hard time.  Which, I'll admit, could mean that I'm wrong (or at least not strictly speaking absolutely right :-).  The counterfactual is, of course, blazingly clear to you, but try as I might, I just can't see the simplicity that you see so clearly, and remain quite convinced of my position.

In the end only math, perhaps, would be capable of closing the gap.  But, barring that, cheers... The much bigger problem is that Jacques Martin is demoralizing and destroying the most brilliant defensive... You're right. At this point that's way more fun!

Oh, "The road to hell is paved with good intentions." But I don't think we can make that excuse for Jacques Martin.

K - I know you won't believe a happy new year is really plausible, but I do wish it to you, with some god-given chance I suppose (and you're right, I don't assume optimality of anything at all, as I don't know of any experiment that could either verify or falsify such a normative concept) Thanks for the proper translation - cheers!


Thank you!  I know you won't believe that happiness is at all meaningful, or even plausibly superior to sadness, but I nevertheless wish you une tres bonne annee!

K - this I'm afraid would take use in a discussion about the meaning of the word *meaning*. With a bit more God-given chance, we'll both be reasonable enough to pass on it. Won't we?

I say they get to the final round this year, and lose it in 6. Would you say I'm being optimistic, or pessimistic? Ok, just forget it - sorry.

The Habs are subject to neither earthly nor metaphysical constraint.  As always, I remain certain of righteous triumph! I'm also certain that we can both let the matter rest on that note.

The comments to this entry are closed.

Search this site

  • Google

Blog powered by Typepad