Picking Losers... - Blog entry - Comments

Update

bruno — Fri, 04 Jan 2008 22:18:42 +0000

Strangely, the second comment (about nuclear and wind) appeared on their website within an hour or two of posting this here. Probably just a cock-up. Still doesn't explain why the first one has never appeared.

Of course if schools were

Vindico — Sun, 30 Sep 2007 12:00:34 +0100

Of course if schools were smart with resources, free from targets, they might conclude that teaching larger classes for some subjects, like History perhaps, and having smaller classes for other subjects like maths might be more appropraite. I am sure maths ability is more graduated thn in History and a clever use of resources would target maths teaching to smaller groups of similar ability.
A voucher system, in which schools were free and independent, would help move towards this.

Aren't Us Taxpayers Generous?!

Vindico — Wed, 19 Sep 2007 10:12:28 +0100

Well isn't that just dandy? Darling has very generously forced all us taxpayer's to stump up OUR money to underwrite the savings of Northern Rock customers. Aren't we generous?

Captcha

bruno — Tue, 18 Sep 2007 12:19:44 +0100

Mark, As you can see, I just tried a couple of tests of the new captcha system (I changed it a couple of days ago after the previous complaints), and it is working straight off the bat for me. How exactly is it not working for you? Are you still getting squares or other unrecognizable characters, or is it not accepting your answer and asking you to try again? One possibility is that it might be a question of case - it's case-sensitive, but to make it easy, all letters now are upper-case. If you have trouble posting a reply, you have my email - let me know privately, and I'll see what I can do to fix it.

Test 2

Anonymous — Tue, 18 Sep 2007 12:15:39 +0100

Another test of captcha

Test

Anonymous — Tue, 18 Sep 2007 12:14:51 +0100

Testing captcha

Judges

Mark Wadsworth — Mon, 17 Sep 2007 20:57:35 +0100

The whole legal profession, CPS, barristers and judges are a totally morally and financially corrupt scam, it's a closed shop run at taxpayers' expense and they can all fuck off if you ask me. A few lay magitrates and a jury can come to better decisions.

(your captcha is really fucking difficult, I am now on my seventh attempt)

Long-term discounting of uncertain outcomes

bruno — Sun, 16 Sep 2007 03:35:00 +0100

Anonymous,

Thank you for the expert criticism. The argument against welfare economics in your first paragraph is key. I would not seek to pretend that this illustration, if it stands up, is any more substantial than any other example. The a priori, deductive arguments against welfare economics, or, from my perspective, scientism in economics more generally, are what count.

The importance of examples is only in their ability to provoke consideration of received wisdom. Although economic studies are probably little better than life in general as a laboratory for experiment into the "science" of human action, there may be an aspect of falsifiability - one cannot prove a contention in this way, but one may help to disprove it. If both A and "Not A" can be argued justifiably from the same proposition, the proposition itself must, as Hegel pointed out, be suspect.

Perhaps Stern and Nordhaus's positions are not precisely diametrically opposed, but they are strongly antagonistic, as Nordhaus himself makes clear. Nordhaus feels he has countered Stern by demonstrating the implausible consequences of Stern's chosen discount-rate. But providing a criticism of Stern's position is not the same thing as proving his own position or disproving Stern's criticism of discount rates such as that preferred by Nordhaus. My contention is that both arguments are "right" in their own terms, and therefore demonstrate the weakness of the proposition on which they are based.

The interesting question is whether, as you doubt, their positions are as contradictory as I, following Nordhaus, suggested. It seems to come down to one's understanding of the purpose of discounting and what it encompasses. It is a fair point that the value of life that is relevant is the value at the time of the disaster, not the value now. And it is also reasonable to assume that people will value lives at least in proportion to their ability to pay, so if they are twice as wealthy on average, lives will be worth at least twice as much on average. But that begs the questions: (a) what do we mean by "growth" and "wealth", and (b) do discount rates conventionally take this into account?

I assume neither of us is taking account of inflation as a monetary phenomenon, which does not represent a growth in wealth, in fact marginally (or significantly in hyper-inflation circumstances) the opposite. I will assume we are working in real, not nominal values, to the extent that it is possible to identify real values.

We are talking instead about the likelihood that, through the division of labour and free trade, more people will be able to afford more goods that they desire than they can now. This, indeed, is a significant part of Nordhaus's thinking, accounting for his argument (to simplify) that we should delay any unnecessary action as much as possible, as we will be more able to afford it in the future - a major factor in his low but rising cost of carbon.

It seems that that argument can cut both ways. Yes, things, including lives, will be more valuable in the future, and we ought to calculate our NPVs on the basis of their value at the time. But equally, we will be more able to afford them in the future, so we ought not to sacrifice more than necessary now when our descendants will be more able to make a larger sacrifice. Is this self-cancelling (at least to some extent), from the perspective of establishing a valid discount rate?

I may be missing a connection, but this seems to me to be a different issue to people's time preference, which is the factor that I understood discounting was traditionally taking into account. I mean by that, very roughly speaking, the economic equivalent of the notion that a bird in the hand is worth two in the bush. The question is whether a discount rate that took account of time preference, and the swings and roundabouts of the likelihood of future growth in terms of present value of future goods, would be a mere 1%, or something more like the Government's 3.5%.

This still seems to me to come down to a judgment call, and one, moreover, that is largely determined by one's view of the desirability of the outcome. If the preferred discount rate is 1% because to use a higher rate would result in unacceptable disregard for the future, then there really isn't much point calculating the NPV of future events based on that discount rate - you have already made the judgment which that calculation is supposed to inform, in order to be able to make the calculation. The same goes for the selection of a higher discount rate that takes an equally (but differently) morally-justifiable stance (that one ought to put a more certain threat to quality-of-life today ahead of a less certain threat to quality-of-life tomorrow). The discount rate is chosen because one has already made a judgement about the relative values of today and tomorrow. It cannot then be used to justify that judgement.

+++++++++

In a way, this is about looking at two sides of two coins. We have two costs in the distant future - one extreme event, and one small but sustained impact. The Stern approach of a minimal discount rate yields a rational result when looking at the extreme event but an irrational result when looking at the small but sustained impact. The Nordhaus approach of a commercial discount rate yields a rational result when looking at the small but sustained impact, but an irrational result when looking at the extreme event. By refocusing on Stern's approach to the extreme event ("a world of nearly-no-discounting-of-life"), and Nordhaus's approach to the small but sustained impact ("a very high price/earnings multiple"), you can argue with justification that their positions are not absurd.

But the other side of the coin does not cease to exist because we are not looking at it. The range of impacts we face do not fall so conveniently under the headings of "extreme" or "small but sustained". If we are going to use the approach of calculating NPVs to inform policy decisions, we are going to have to choose a figure that will apply to all impacts, small or extreme, short or sustained, and everything in between. If we choose a low discount-rate, we will have to consider the other side of the small-but-sustained coin, where the action that is mandated today is excessive relative to the avoided harm. If we choose a higher discount-rate, we will have to consider the other side of the extreme coin, where deaths sufficiently distant in the future will be excessively devalued in today's terms. From what I can see, your justification of the two approaches is a fair defence of the light sides of the coins, but they do not refute the unavoidable reality of the dark sides of both coins, as highlighted by each protagonist when criticizing the other perspective.

Discounting is a useful financial tool for indicating whether an investment is likely to be worthwhile over a medium-term (say ten- to twenty-year) period. With sufficient expertise and insight, future values can be estimated over such a period with a reasonable hope of being approximately accurate, and the impact of discounting will not make the accuracy of such judgments irrelevant over such a limited period (though it is getting that way on a 20-year DCF analysis). But, in my opinion, discounting and NPVs are wholly inappropriate tools to use for estimating long-term values over many decades or centuries, where future values are increasingly uncertain as the period extends, and yet the effect of variations in those values is diminished by the compound effect of discounting over such a long period. If you have to use this approach to estimate NPVs, then of course you try to tailor the discount rate to achieve what appears to be a rational outcome according to your moral preferences and assumptions about the future. But you cannot then use calculations based on the discount-rate derived in that manner in order to justify arguments about the right moral course of action based on the scale of future costs - that is simply circular reasoning. In which case, what exactly did the mathematical part of this process gain you?

For such uncertain and long-term impacts, a different approach is needed. Unfortunately, mathematical economics needs some method to calculate future values. Any alternative mathematical approach will suffer from similar problems. Mathematical economics simply does not provide a useful tool to derive policy prescriptions for an issue like AGW. I believe this is symptomatic of a broader malaise with the approach, and a broader ineffectiveness for policy guidance, but I raise it only as an unusually accessible example of its failings, not as a falsification of the principle of mathematical economics. That comes from the sort of deductive analysis of the nature of value that you describe. But my hope is that, as an illustration of the weakness of the approach, it may cause one or two people to consider whether something that is so abstracted from reality, and that can provide such different answers from small, subjective changes in conditions, really provides a sound basis for policy prescriptions.

You sound like an economist. If so (or even if not), if you get the chance, read Nordhaus's section on Stern and discounting (section IX, pp.137-167), which explains the issues much better than I can. For instance, I have simplified by referring only to the discount rate, when Nordhaus is at pains to emphasise that this isn't just a question of the discount rate, but also of the utility function - the other aspect of the social welfare function. The discussion of the complementarity of the discount rate and the utility function addresses some of your arguments with regard to the justifiability of a 1% (net) discount rate.

Mathematical economics

bruno — Sat, 15 Sep 2007 20:53:00 +0100

Fake consultant,

Sorry for the delay.

I was criticizing mathematical economics specifically, not maths generally, nor its application to climate-change science. Climate-modelling ≠ economics. Oceonography ≠ economics. Atmospheric physics ≠ economics. In these and other areas of science, I support further study to improve our understanding of the relevant systems.

The anonymous commenter after you has given the broader reasons to be critical of the notion of welfare economics (of which environmental economics is a branch). I will respond to that erudite submission separately, but the general point they make about "quantifications of things that cannot or should not be quantified" is the real basis of the argument. I was simply attempting to use a clash of two welfare economists to illustrate the difficulties.

It is not clear to me how you think further studies will help to clarify the appropriate discount rate to use for valuing present costs of future impacts. Even as we refine the science and pin down the risk of those impacts with greater accuracy, a difference of one or two per cent in the discount rate chosen for the purposes of valuing those impacts will have a dramatic impact on the policy prescriptions that follow. As Stern and Nordhaus have demonstrated, it's the difference between valuing current carbon-emissions at $30 or $300 per tonne.

Unfortunately, economics is not amenable to the same approach as the natural sciences. There is a big jump from understanding what might happen to our climate if people behave in a certain way, to the self-referential question of predicting people's behaviour and how incentives might affect that behaviour. The problem with the attitude of a lot of scientists, politicians, journalists, and other intellectuals (in the Hayekian sense) is that they act as though the policy prescriptions flow naturally from their scientific study of climate systems. The questions regarding human behaviour that follow from the science are actually more complex than the climate science itself.

On a couple of points of detail:

1. Is CO2 a "leading indicator of global mean temperature rise"? My understanding is that, historically, it has been a trailing indicator.

2. The date of 2200 is an entirely arbitrary one selected by Nordhaus. His point is only reinforced by consideration of the difficulty of pinning down the date and scale of future impacts with any accuracy at this time. As many of those impacts are dependent substantially on human behaviour, this is inherently unpredictable, even as our scientific understanding of the consequences of human behaviour improves.

Captcha

bruno — Sat, 15 Sep 2007 19:49:50 +0100

Thanks to both of you for pointing out the captcha problem. I've just spent a good part of a day trying and failing to figure out what is causing it. Fortunately, there are alternative modules, so I've now replaced the module we were using with another one that seems to work OK for me. If you still get problems, and it's bad enough that you can't submit a comment, you can contact me by my contact form at http://www.pickinglosers.com/user/2/contact. I'm sorry we have to use captchas for anonymous users, but we were getting terrible problems with comment spam. You are welcome to register as a user of the site. Authenticated users are not asked to fill out a captcha form with each comment.

Too harsh on welfare economists

Anonymous — Fri, 14 Sep 2007 07:23:31 +0100

I agree with the feeling that neo-classical welfare economics is a sham because of its attempts at basing policy prescription on quantifications of things that cannot or should not be quantified for that purpose.

But I don't think that either of the two examples you quote are absurd enough to make that point. I did not have the time to trace whether either Stern or Nordhaus original examples are vulnerable to my critique, so I am only responding to your formulation.

Using an ordinary discount rate (like the Government's current 3.5 per cent) does not mean that "it isn't worth doing much now to avoid large numbers of deaths". Any sane person applying cost-benefit analysis (okay, contradiction in terms, but nevermind) to a long-term problem would need to consider the specific inflation rates applied to the values of different things. When it comes to avoiding deaths, the respectable hypothesis is that willingness to pay grows in line with income or consumption (or even faster if you believe some academic research). Make that 2.5 per cent a year above the same inflation rate as the one used in defining the 3.5 per cent Government discount rate, and it is now worth doing quite a lot to avoid future deaths: the effective discount rate on life is "only" 1 per cent. Add to the mix the theories in the Green Book about lower discount rates for longer time period and you have reached a world of nearly-no-discounting-of-life from a standard discount rate method. Where is the absurdity?

As for the example attributed to Nordhaus, there seems to be a similar interaction between economic growth and discounting. If you had an investment opportunity that would generate profits of 0.1 per cent of global consumption from 2200 and forever more, guaranteed, what would the shares in that investment be worth today according to a standard financial method? Let's look at it with a dividend growth model. This is a guaranteed opportunity with guaranteed income growth so we will be getting most of our cost of capital covered by the dividend growth element, and we can affort a very high price/earnings multiple. I haven't done the numbers but investing the equivalent of half a year's consumption does not seem manifestly absurd in exchange for a small but guaranteed, growth- and inflation-proofed, permanent income stream.

Now plainly it would be daft to cut consumption by 56 per cent in 2008 on the foot of some global warming theory. But don't accuse the discounting maths of that bit of silliness: the real reason why it would be daft is that there does not exist an opportunity to spend $30 gazillions tomorrow to achieve the 0.1 per cent guaranteed extra consumption from 2200 forever more.

(And your captcha gives me squares too.)

Govt consultation

Mark Wadsworth — Sat, 08 Sep 2007 08:29:04 +0100

Out of interest, whose response to what bit of consultation?

And why do you 'captchas' nearly always have squares in them? What letter is that? It usually takes four or five attempts to get one with just letters.

Dave is totally mad

Mark Wadsworth — Fri, 07 Sep 2007 21:55:04 +0100

Even better, i) he is recycling a policy he first suggested in early 2006 and ii) it is going to be voluntary, so our broken society will fix itself by itself!

with all respect...

fake consultant — Fri, 07 Sep 2007 09:20:53 +0100

...come of your comments seem to require a bit of reconsideration.

for example, the idea that mathematics is somehow now discredited because modelling cannot, at this moment, provide exact answers seems to be an unreasonable conclusion.

remember that input factors do become more accurate as research teaches us more about the behavior of the subject we're modelling.

an example is found in oceanography. the accuracy of models that predict the behavior of surface objects are constantly improving. does that mean mathematics as a tool was discredited until the models became more accurate, or did the inputs improve, making the underlying math process look better, or does a combination of both occur?

the answer is both. that tells us that the process of using mathematics to examine potential results remains sound, even as the math improves.

additionally, you do not seem to have considered the possibility that 2200 might be a date that itself is subject to change.

here's an example from a scientific consortium that includes the us national science foundation, the universities of calgary and colorado, the us geological survey, and pennsylvania state university that tells us a rise in sea level of 13 to 20 feet has occurred before-130,000 years ago-from a similar amount of glacial melt as is currently modelled to occur by 2100.

the modelling data was corellated to ice core data. what does that accomplish? ice cores contain air samples trapped at the time the ice froze, and that air can be analyzed for carbon dioxide content.

co2 is a leading indicator of global mean temperature rise, and the ice core record proves today's co2 levels are as high as they were 130,000 years ago.

the combination of the two is a very strong indicator that warming is accelrating at a rate we had not expected, and it is this potential for accelerated warming that also seems to be lacking in your consideration.

it is relatively easy...

fake consultant — Fri, 07 Sep 2007 08:46:01 +0100

...to look only at the money spent vs. results obtained to try to understand what went wrong or right with this expenditure-but that would be a misleading exercise.

for example, if the entire amount went to repair or replace crumbling buildings, there would be no reason to expect a substantial return on test scores.

on the other hand, if the entire amount was spent on academic initiatives with the same lack of results, your displeasure would be well justified.

it would be very interesting to see you run with this story, and it seems to be crying out for more specific analysis.