I just started reading the Machinery of Freedom by David Friedman. I find the beginning section very well made and convincing. As in this is easily material that could be given to an intellectual person and have him understand it and possibly agree with it. Perhaps some buttering up beforehand would help, but the book (besides its too-early mention of anarcho-capitalism) is pretty solid and gives nice utilitarian arguments for AnCap (even for a semi-deontologist like me).
Thoughts?
Oh, and in case Friedman sees this post with his uberbots crawling the web, thanks! (though I disagree with you on utility comparisons :P )
I enjoyed one of his novels and a couple of his short articles, but I've never read any of his nonfiction books. Sounds like it's about time to learn what he has to say. So far I like it very much. Some other people have written similar observations, but he is very observant and has a great writing style.
The writing is certainly extremely easy to read. Yet the more I read of his mainstream welfare economics approach in the book the more ticked off I am. It goes against the foundations of Austrian thought.
Either way, the rest of the book is great.
"Oh, and in case Friedman sees this post with his uberbots crawling the web, thanks! (though I disagree with you on utility comparisons"
You think interpersonal utility is entirely incomparable? You have no opinion at all as to whether you would suffer more disutility from being tortured to death than I would from a cold shower?
"It goes against the foundations of Austrian thought."
You might almost suspect that I'm not an Austrian.
Isn't the more relevant question whether it's right?
@David Friedman
Are you planning on writing a sequel to Salamander? Sorry for the off-topic post.
You have no opinion at all as to whether you would suffer more disutility from being tortured to death than I would from a cold shower?
But isn't this better understood as an intuitive theory of human nature?
One of the essential things about choosing is that my choices happen in my head. Hence, if a sadist offers you the alternative to suffer a cold shower or watch me be tortured to death, it is not literally the expectation of my suffering that impels your choice but, rather, your own anticipated feelings of sympathetic pain, lifelong guilt, and failure to fulfill your own moral self-image that impels you to undergo the suffering of a cold shower to spare me. And this account of choice should not be understood in any normative sense, it is merely descriptive of the actual causal factors that drive your choice. It is a matter of fact that you will not feel my torture. Only I will feel it because only I am me. Hence, there is a significant difference between choosing whether to undergo a shower or watch me be tortured to death and choosing whether to undergo a shower or be tortured to death yourself.
Clayton -
I've written about the first 2/3 of the sequel, but have been stalled at that point for quite a while.
One of the protagonists is Eirick, Lord Iolen's son, stranded in Forstmark at about age 10 by the death of his father. As Mari remarks at one point in the book, he is a much nicer person than his father.
I liked Machinery of Freedom a lot.
You might almost suspect that I'm not an Austrian. Isn't the more relevant question whether it's right?
Well, yes, I know you're not Austrian :P
You cannot compare the utilities of other people. When you have to make the decision between eating an apple and giving it to a homeless person you compare your expectation of their utility (given your own utility for apples and your understanding of their tastes) discounted by how much you care about them against your utility gained from eating an apple. [Edit: this is unclear. Let me clarify: the imporant thing here is the discount, not the expectation, so maybe the beginning of the sentence should be inverted - you're comparing how much you care about their utilities to how much you care about your own utilites, and the care for their utility is informed by their possible reaction to being given an apple]
Wouldn't you say that making external comparisons of utilities relies on cardinal utilities? Internal comparisons rely on a natural funciton of the brain which creates our ordinal utilities. Even if we could have a function which gives us the exact utlity orderings of other people, we would still be left with two ranked but non-cardinal schedules that would need to be merged, which is even theoretically impossible.
Then you're left with the option of accepting cardinal utility calculation. This means that we need to define a 0 utility and a +1 increase in utility. I am open to suggestions where you want to define the 0 and what a +1 increase means. Furthermore, I highly suspect that utility itself is not a value that can be added up over time and hence be accumulated. Instead, it's merely a function that exists at the moment to help make decisions based on satisfaction and expected satisfaction.
To return to your original example,
You have no opinion at all as to whether you would suffer more disutility from being tortured to death than I would from a cold shower
I have a very strong opinion about my utility if I have to be tortured to death and my utility from choosing for you to take a cold shower instead.
I found it to be a fairly easy book to follow and it didn't have much of what I'd consider to be moral preaching which I enjoyed.
"Wouldn't you say that making external comparisons of utilities relies on cardinal utilities?"
No--although I have no objection to cardinal utilities.
"Internal comparisons rely on a natural funciton of the brain which creates our ordinal utilities. Even if we could have a function which gives us the exact utlity orderings of other people, we would still be left with two ranked but non-cardinal schedules that would need to be merged, which is even theoretically impossible.
Then you're left with the option of accepting cardinal utility calculation. This means that we need to define a 0 utility and a +1 increase in utility. I am open to suggestions where you want to define the 0 and what a +1 increase means."
You might want to look at the work of John Von Neumann, who solved that particular problem quite a long time ago. He showed that if individual choices under uncertainty met some fairly simple consistency conditions, cardinal utility was implied by choices up to linear transformations.
That doesn't say where the zero point is--I think there is much to be said for defining it as the suicide point. And it doesn't say what the unit is--but then, units are essentially arbitrary for most cardinal measures, no reason why Centigrade is more natural than Fahrenheit, or vice versa. But the statement "I value A more than B by twice as much as I value A more than C" turns out to be a meaningful statement, whose truth or falsity can be revealed by choices.
On the final point, I repeat my question, since you didn't answer it. Do you have no opinion as to whether your disutility from being tortured to death is more or less than my disutility from a cold shower? If you prefer, make it two other people. Are you agnostic as to whether, on average, people whose last name starts with B get more disutlity from being tortured to death than people whose last name starts with C get from taking a cold chower?
David Friedman:You think interpersonal utility is entirely incomparable? You have no opinion at all as to whether you would suffer more disutility from being tortured to death than I would from a cold shower?
Interpersonal utility seems to be entirely incomparable in an objective (mathematical/quantitative) sense. That in no way means I have no opinion at all as to e.g. whether I'd suffer more disutility from being tortured to death than you would from a cold shower. Opinions aren't facts. It's certainly possible for you to suffer more disutility from a cold shower than I would suffer from being tortured to death.
The keyboard is mightier than the gun.
Non parit potestas ipsius auctoritatem.
Voluntaryism Forum
Do you have no opinion as to whether your disutility from being tortured to death is more or less than my disutility from a cold shower?
That is correct.
How can you merge two ordered lists that don't have some basic common unit of ordering?
It's like taking my preference for movies (1-> X, 2-> Y 3-> Z) and yours (1->R, 2-> Q, 3-> S) and merging them without considering some value like "rating out of five stars". There is no theoretical way of merging the above lists just as ordered lists.
John Von Neumann
I'll look into it.
I think there is much to be said for defining it as the suicide point
If 0 is the suicide point, then there is nothing worse than suicide. That means that having your arm chopped off and then committing suicide is better than committing suicide. No?
but then, units are essentially arbitrary for most cardinal measures, no reason why Centigrade is more natural than Fahrenheit, or vice versa.
You still need a 0, otherwise you can't say "twice as much." When people only had C and F they could not say "twice as hot" in a physical sense. Only when K came around could they do that.
But the statement "I value A more than B by twice as much as I value A more than C" turns out to be a meaningful statement, whose truth or falsity can be revealed by choices.
Could you give me an example?
Mr. Friedman, would Theory of Games and Economic Behavior cover the topic adequately?
Hm, in the book they themselves say that it is too short to give a good discussion of utilities... Any papers instead of this book?
"If 0 is the suicide point, then there is nothing worse than suicide. That means that having your arm chopped off and then committing suicide is better than committing suicide. No?"
No. There are numbers lower than zero. Lots of them.
"You still need a 0, otherwise you can't say "twice as much."
Consider temperature. I can say that the increase in temperature from yesterday to today is twice the increase for the same dates last year, even if the zero point is arbitrary. Similarly here. We are always comparing alternatives, and I can say that alternative A is better than B by twice as much as C is better than D--no zero point required.
"Could you give me an example."
Von Neumann solved the problem by considering choices under uncertainty. Suppose I am indifferent between a certainty of outcome B and a coin flip between A and C. That means that [U(A)+U(C)]/2 = U(B), where U is Von Neumann utility.
My summary of the ideas is in:
http://www.daviddfriedman.com/Academic/Price_Theory/PThy_Chapter_13/PThy_Chapter_13.html
Go down to the optional section at the end, "Choice in an Uncertain World."
@Wheylous
I think that the concept of lotteries which is used to construct Von Neumann utility is a kind of thought-experiment because we rarely, if ever, actually know the probabilities of the respective outcomes of our choices nor are payouts guaranteed. Hence, VN utility exists in the same kind of state as an Evenly-Rotating Economy. Both describe a counter-factual world which may be helpful in illuminating the real world as long the conditions imposed on our conclusions by the thought-experiment are not overlooked. VNM speaks of "choice" but this is not real choice, it is a kind of idealized choice that may or may not share essential characeristics with real choices.
Austrian economists also maintain that utility is ordinal. However, they challenge the use of mathematical utility functions by neoclassical economists on the grounds that such functions yield cardinal utilities, “measured,” usually, in utils.3 Neoclassicals respond by asserting that, in dealing with bundles of goods: (1) a function that ranks bundles in accordance with an individual’s necessarily ordinal preference ranking is an ordinal function, and the ranking it generates is ordinal; (2) because the ranking of bundles generated by a specific utility function (F) remains the same after any positive monotonic transformation into another function (G), G, also, is a utility function; and, (3) in that case, it does not make any difference whether an individual’s preferences are represented by F, by G, or by any other function that is a positive monotonic transformation of F (or of G for that matter).4 However, maintaining that “properly constructed” utility functions are valid representations of, admittedly, necessarily ordinal utility is very misleading, and confuses and obfuscates the issue, because as these functions are to be the objects of calculus operations they necessarily require cardinal numbers.5 Thus, the use of such functions necessarily implies that we are simultaneously in the realms of cardinal and ordinal utility,6 which is problematical, at best. The problem originates in their interpretation of utility functions as ordinal and not cardinal. It is compounded by two other factors: the assumption that utility functions are differentiable; and the failure to use, correctly and consistently, dimensions in their work with utility functions. It is my contention that their analysis is incorrect, both as a matter of praxeology and as a matter of mathematics.
Where there is no property there is no justice; a proposition as certain as any demonstration in Euclid
Fools! not to see that what they madly desire would be a calamity to them as no hands but their own could bring
@nir: Great article.
For neoclassical economists the existence of these “ordinal” utility functions makes possible a virtual utopia—they can have their cake and eat it too. They get the truth of subjective valuation in ordinal utility rankings, and, also, the supposed objectivity of cardinal number rankings, which allows them to use calculus. But there can be only be an ordinal ranking of utilities. That is, the only thing that may be said in comparing two bundles is which is preferred to the other or whether neither is preferred to the other.15 That is the crux of the matter. Neoclassical economists act as if they can have both cardinal and ordinal rankings at the same time and in the same respect. That is, they say that because they can order bundles on the basis of the cardinal number assigned to it by a specific function, the order so generated is a rank-order, and that the utilities being so considered are ordinal. This is incorrect. Were it truly an ordinal ranking, it would not be absolutely necessary to use the cardinal numbers from which the ranking was generated in their mathematical calculations. Rather, the ordinal numbers corresponding to the ranking generated from the cardinal numbers could themselves be used in such calculations. Therefore, the order generated by neoclassical utility functions is not a rank ordering in any meaningful sense of the word.
For neoclassical economists the existence of these “ordinal” utility functions makes possible a virtual utopia—they can have their cake and eat it too.
They get the truth of subjective valuation in ordinal utility rankings, and, also, the supposed objectivity of cardinal number rankings, which allows them to use calculus. But there can be only be an ordinal ranking of utilities. That is, the only thing that may be said in comparing two bundles is which is preferred to the other or whether neither is preferred to the other.15 That is the crux of the matter. Neoclassical economists act as if they can have both cardinal and ordinal rankings at the same time and in the same respect. That is, they say that because they can order bundles on the basis of the cardinal number assigned to it by a specific function, the order so generated is a rank-order, and that the utilities being so considered are ordinal. This is incorrect. Were it truly an ordinal ranking, it would not be absolutely necessary to use the cardinal numbers from which the ranking was generated in their mathematical calculations. Rather, the ordinal numbers corresponding to the ranking generated from the cardinal numbers could themselves be used in such calculations. Therefore, the order generated by neoclassical utility functions is not a rank ordering in any meaningful sense of the word.
Earlier, he states:
Let us be quite clear. Neoclassical economists wish to apply calculus to utility functions; i.e., they wish to take their derivatives. In fact, that is the sole purpose for restating preference rankings in the mathematical language of functions. However, for calculus to be applied to these functions it is necessary that the numbers assigned be cardinal.
utility functions; i.e., they wish to take their derivatives. In fact, that is the sole purpose for restating preference rankings in the mathematical language of functions. However, for calculus to be applied to these functions it is necessary that the numbers assigned be cardinal.
[Emphasis added]
This quote throws light on why Friedman's torture-me-versus-shower-you thought-experiment doesn't get us cardinal utilities that are useful beyond the ordinal rankings they are constructed to represent - calculus is also known as the infinitesimal calculus... if you wiggle x by an infinitesimally small amount, what is the ratio by which y changes with respect to x? While it is undeniable that the human brain can and does perform interpersonal utility comparisons at a gross level (particularly in forming moral judgments), there is no way to get from there to infinitesimal variations in utility, which is what is implied in a real-valued cardinal utility suitable for use with the tools of calculus.
I'm not sure if that necessarily makes any notion of cardinal utility useless for economics, I think it boils down to specifying the limitations of the counter-factual model in representing actual human behavior.
"Austrian economists also maintain that utility is ordinal."
That "also," and other things you say, imply that you believe that neoclassical economists maintain that utility is ordinal. That was true a hundred years ago, and it may describe how some textbooks present the subject, but it hasn't been an accurate description of neoclassical theory in general since Von Neumann showed how to cardinalize utility--as I already pointed out in this discussion.
Note that diminishing marginal utility of income only makes sense if utility is cardinal--and is demonstrated by the existence of the insurance industry.
Note that diminishing marginal utility of income only makes sense if utility is cardinal
How so?
"How so?"
Diminishing MUI means that the difference between my utility for $10,001 and my utility for $10,000 is more than the difference for $20,000 and $20,001. But if utility is only ordinal, not cardinal, that's meaningless. All one can say is that I prefer $20,001 to $20,000 and $10,001 to $10,000, but not by how much.
Mr. Friedman, may I ask what is your opinion about the Austrian business cycle theory? You don't have to give me an extensive answer. I would just like to know how many serious libertarians outside the Austrian school of economics actually accept it as a fact. I thank you for your answer beforehand. (I hope you can understand my English.)
And I would like to present my apologizes to the original poster. I don't want to cause offtopic and ruin your thread.
I don't work in macro, so don't really have an opinion of my own. My father's view, as best I understood it, was that the theory was logically possible but inconsistent with the available evidence.
Diminishing MUI means that the difference between my utility for $10,001 and my utility for $10,000 is more than the difference for $20,000 and $20,001
Does it? You could have a utility function of y=-x2 + 10
Consider only the first quadrant. DMU doesn't say anything about second derivative (if, for a moment, we assume continuity and differentiability), just the first.
It almost beats For A New Liberty imo. David Friedman is brilliant.
Marginal utility of income is the first derivative of total utility wrt income. Diminishing marginal utility of income means that marginal utility decreases with income, hence that the second derivative of total utility wrt income is negative.
It doesn't have to be true--one can obviously write utility functions for which it isn't. But that is what it means, and there is good reason to think it usually true. And it corresponds to risk aversion wrt income, which is also usually true.
And it doesn't mean anything if total utility is only ordinal, not cardinal, for the reason I already pointed out.
David, the problem I see is that you define Diminishing Marginal Utility in an overly technical mathematical way, no doubt because maths is the way in which neo-classicals like to do their economics. However this is not necessary for economic analysis, and the Austrain method works quite well with ordinal value and a meaningful diminishing marginal utility (Though of course utils don't come into it).
See Bohm-Bawerks exploration of the topic, wherein he imagines the varying purposes to which a pioneer farmers several sacks of grain could be put. The loss of any one would impact on the achievement of his least valued end. This is the diminishing marginal utility. His most valued ends are met first, additional sacks are allocated to ends that are valued such that they are lower down the ordinal scale.
There is certainly nothing 'meaningless' about this; however we concede that mathematical statements of 'but by how much less valued' are jettisoned. But we repeat ourselves.
@nir: "an overly technical mathematical way"
While I agree that the real-valued utility function is probably much less useful than mainstreamers think it is, I think we have to be careful with this broad brush of "overly technical" because many Austrian arguments - particularly ABCT which is one of Austrian economics' most important results - are, in fact, very technical. If we say "yeah but they don't use maths", the natural question is "so what?" What is wrong with mathematics? If you have five people in a room, then five more enter the room, are there not ten people in the room? Does not this arithmetic fact impinge on praxeology as much as any law of logic?
I think the real issue is that the mainstream economists do not properly circumscribe the generality of their conclusions. If you look up the definition of Von Neumann utility which Prof. Friedman has mentioned, you will see that is defined in terms of thought-experimental "lotteries" in which a "choice" is made between a bundle of goods with probability p and another bundle of goods with probability 1-p. Immediately, we can see two counterfactual aspects to this thought-experiment: actual probabilities are rarely known beforehand (outside of a casino) and payouts are never absolutely certain. This means that significant qualification of conclusions drawn on the basis of VN utility thought-experiments is required. The conclusions of such analyses only hold to the extent that not knowing probabilities beforehand doesn't matter and not being absolutely certain of payment doesn't matter. I think that, most of the time, both of those conditions matter a lot.
I try to answer this question:
David Friedman:On the final point, I repeat my question, since you didn't answer it. Do you have no opinion as to whether your disutility from being tortured to death is more or less than my disutility from a cold shower? If you prefer, make it two other people. Are you agnostic as to whether, on average, people whose last name starts with B get more disutlity from being tortured to death than people whose last name starts with C get from taking a cold chower?
I think there is an important implication in this question, which is that somebody forces you or someone else to be tortured or take a cold shower, in which case we can use intuition and experience that very few people freely choose to be tortured while a lot more take cold shows. Hence we could conclude in this case that on the average for most people torturing will be worse, and we can make an interpersonal utility comparison and decide on it. So if we were forced to decide if someone should be forced to be tortured or given a cold shower (no other alternatives available), I guess the answer was obvious.
Yet if people voluntarily chose to do one or the other this is not possible anymore. The utility that people derive from torturing themselves because of sexual desires or religious believes or whatever cannot be compared with people who like to take cold showers.
That is where everything goes wrong. People believe that redistributing wealth was good on the basis of interpersonal utility comparisons because they think high or low income is a thing that happens to you or not independent from your own value judgments and choices which is of course not true at all. And if not more important of course because they are ignorant how wealth is created in the first place, they think it is there or not there anyway, it is only a matter of distribution..
I am not 100% sure but for my point of view the claim "interpersonal utility comparisons are not possible at all" is an overstatement. You would need to add the qualifier "in case only voluntary chosen ends are pursued".
@ Wheylous,
Yeah, I liked MOF quite a lot, and if it were available in German I would have given it away multiple times already... Unfortunately it is not, or is it, or is there something planned?
@ David
I have a question to Von Neumann. If I understand this correctly this method would only allow you for saying things like:
- Bob likes 1 apple twice as much as 4 oranges,
- John likes 1 apple twice as much as 2 oranges
Yet those are relative statements on two different and independent value scales. Both value scales have still no connection to each other, because it doesn’t reveal if Bob and John like one apple to the same degree, or the orange to the same degree. Is that correct?
BTW: I knew Von Neumann (I guess a different one) only from computer architecture
nirgrahamUK:David, the problem I see is that you define Diminishing Marginal Utility in an overly technical mathematical way, no doubt because maths is the way in which neo-classicals like to do their economics. However this is not necessary for economic analysis, and the Austrain method works quite well with ordinal value and a meaningful diminishing marginal utility (Though of course utils don't come into it). See Bohm-Bawerks exploration of the topic, wherein he imagines the varying purposes to which a pioneer farmers several sacks of grain could be put. The loss of any one would impact on the achievement of his least valued end. This is the diminishing marginal utility. His most valued ends are met first, additional sacks are allocated to ends that are valued such that they are lower down the ordinal scale. There is certainly nothing 'meaningless' about this; however we concede that mathematical statements of 'but by how much less valued' are jettisoned. But we repeat ourselves.
I'm kinda playing Devil's advocate here, but couldn't you say that the ratios between the numbers of sacks allocated to the different ends shows how much more the pioneer farmer values one end over another?
Autolykos,
Thanks for the question.
lets say you have 5 ends the attainment of which each require 1 sack. These are a,b,c,d,e
a is feeding your prize cow for the month,
b is planting crops for next year
... skip a few...
e is feeding your pet parrot for the year
As I hand you a sack, you allocate first to a. I give you another sack, you allocate to b.
By the time you have 5 sacks and are fully allocated, if I should steal a sack at random, the loss you incur is the loss of e, since with reallocation of your remaining sacks, you can still acheive a,b,c,d.
Now what are the ratio's of sack allocation between uses. They are 1 to 1. This says nothing about how you quantitively value the attainment of your ends...
the ratios between the numbers of sacks allocated to the different ends shows how much more the pioneer farmer values one end over another
So by paying $30 per month for internet you show you like internt 6 times more than $5 footlongs at Subway?
"I am not 100% sure but for my point of view the claim "interpersonal utility comparisons are not possible at all" is an overstatement. You would need to add the qualifier "in case only voluntary chosen ends are pursued"."
It's still an overstatement. We can observe how much trouble people are willing to go to in order to pursue ends voluntarily chosen, and form some opinion from that about how much utility they get from them.
More generally, in understanding human beings we have the advantage of ourselves being human beings. So although there is no precise way of measuring the utility other people get from things, we can get at least some information about it by observing their behavior, by introspection, and in a variety of other ways.
A German translation of Machinery was published in 2003--I don't know if it is still in print or not. Unlike the French translation, which I only got to look at when it was done (and sent off an anguished letter pointing out multiple mistakes), for the German one I was able to interact with the translator, so I think it is reasonably accurate, subject to the limitations of my not very fluent German. For similar reasons, the German translation of my Hidden Order is the one translation of that book which I'm pretty sure is accurate.
For some reason publishers don't encourage interaction between authors and translators.
1. It's the same Von Neuman--one of the more impressive intellects of the 20th century.
A Nobel prize winning physicist is quoted as saying "There are two kinds of people. John Von Neumann and the rest of us."
And Fermi's comment after meeting Von Neumann, to one of his colleagues, was "That man is cleverer than I am. Just as I am cleverer than you are."
2. Von Neumann utility isn't about interpersonal comparisons--it's a way of making utility cardinal rather than just ordinal. So your basic point is correct.