Is it better to tax consumption or to tax income? Is it better to tax carbon or to mandate fuel efficiency? Is it better to foster global competition or to protect local industries?
Today, I will attack none of these questions. Instead, I will attack the meta-question of how to attack such questions. For economists evaluating alternative policies, the industry standard is the efficiency criterion, also known as the welfare criterion. (I’ll illustrate what that means as I go along.) But now comes Princeton Professor Uwe Reinhardt with a piece in the New York Times that questions the orthodox approach found in virtually all modern textbooks (including one in particular).
Let’s first dispense with the straw man. I’ve never heard of an economist who believes that every efficient policy is good, and I’ve heard of very few who believe that every inefficient policy is bad. It’s true that most economists do seem to believe that any good policy analysis should start by considering efficiency. That doesn’t mean it should end there.
I think economists are right to emphasize efficiency, and I think so for (at least) two reasons. First, emphasizing efficiency forces us to concentrate on the most important problems. Second, emphasizing efficiency forces us to be honest about our goals.
I’ll illustrate the first advantage with a stylized example adapted from Chapter 17 of The Big Questions. Suppose you live next door to Bill Gates. Bill likes to play loud music at night. You’re a light sleeper. Should he be forced to turn down the volume?
An efficiency analysis would begin, in principle (though it might not be so easy in practice) by asking how much Bill’s music is worth to him (let’s say we somehow know that the answer is $10,000) and how much your sleep is worth to you (let’s say $25). It is important to realize from the outset that no economist thinks those numbers in any way measure Bill’s subjective enjoyment of his music or your subjective annoyance. Only a crazy person would think such a thing, and I’ve never met anybody who’s that crazy in that particular way. Instead, these numbers primarily reflect the fact that Bill is a whole lot richer than you are. Nevertheless, the economist will surely declare it inefficient to take $10,000 worth of enjoyment from Bill in order to give you $25 worth of sleep. We call that a $9,975 deadweight loss.
Why is that an important calculation? Here’s why: It reminds us that there might be a better solution to this problem, such as allowing Bill to crank up his speakers and forcing him to pay you, say, $5000 in compensation. Compared to shutting down the music, that’s better for Bill and better for you. Maybe that’s an alternative we should consider. In fact, whenever a policy is inefficient, there’s always an alternative policy that, in principle, is better for everyone. That’s what inefficiency means.
Now in this case the proposed alternative policy might not be such a good idea. It might, for example, encourage Bill to lie about the value of his music, and encourage you to lie about the value of your sleep. It might even encourage you to move a little closer to Bill just so you can find more things to complain about and get compensated for. Or it might have negative long-term consequences for the way we think about wealth and social status. So maybe we don’t want to pursue this alternative policy after all. But, says the economist, we ought at least to consider it.
And — here’s the point — the bigger the deadweight loss, the greater the potential gains from an alternative policy. Therefore, the bigger the deadweight loss, the more it’s worth at least attempting to devise a good alternative policy. We calculate the deadweight loss as a rough but useful guide to how much effort we should put into this problem. (Calculating deadweight losses is the same thing as worrying about efficiency.)
Take a more realistic example: Should we spend, say, a billion dollars a year to subsidize end-of-life health care for poor people? It would be, I think, a terrible mistake to settle this question without at least asking whether the recipients might prefer that we spend our billion dollars some other way — say by subsidizing their groceries or just giving them cash. If so, the difference in value between what they’re getting and what they could be getting (as measured by the recipients) is a deadweight loss. The bigger that deadweight loss, the more we should reconsider our spending priorities.
Now once again, efficiency is not the be-all and end-all of policy analysis. Even if poor people prefer subsidized groceries to subsidized health care, we might still choose to give them health care if, for example, we believe that they are less likely to make wise decisions about their own health care than about their own groceries, or if we’re afraid that subsidizing groceries (or handing out cash) is somehow more likely to invite fraud. But the bigger the deadweight loss, the more we should probably rethink those concerns, because the bigger the deadweight loss, the more opportunity there is to improve life for the recipients. That’s the first reason we should care about efficiency.
The second reason we should care about efficiency is that efficiency analysis strikes down political smokescreens. Like this:
Politician: Here’s my program to make the health care system work better by subsidizing health care for the poor.
Economist: Your program costs a billion dollars and delivers half a billion dollars worth of benefits. That’s inefficient.
Politician: So what?
Economist: Well, the “so what” is that maybe you could take that billion dollars and deliver a full billion dollars worth of benefits instead if you spent it a little differently. Why not just hand the cash out to poor people?
Politician: Because I don’t want to help all poor people. I only want to help sick poor people — and this is the only way I can think of to do that.
Economist: Ah. So your goal here is not to make the health care system work better after all. Instead it’s to transfer resources to sick poor people.
Politician: I guess so.
Economist: That’s fine. Now we can have a healthy debate about whether that’s what we want to do.
And now, you see, thanks to the economist’s insistence on thinking about efficiency, we end up having an honest debate about the politician’s real goal instead of a dishonest debate about the politician’s feigned goal. However the debate turns out, that’s a useful exercise.
It’s not just politicians who sometimes hide their true goals behind smokescreens. Suppose, for example, that one of Professor Reinhardt’s colleagues were to write a series of columns in the New York Times calling for more fiscal stimulus, including higher unemployment benefits. On some days, he argues that these policies will increase GDP. Other days, he argues that they will reduce unemployment. Other days, he tells you that it’s cruel to deny benefits to suffering jobseekers.
Those are of course all different (though intertwined) arguments. You might accept some but not others. Even if you accept them all, you still can’t draw a policy conclusion until you weigh these benefits against the policies’ offsetting costs (including the opportunity cost of the expenditures, the effect on long run growth, and so forth). The advantage of an efficiency analysis (along, say, the lines suggested here) is that it would force Professor Reinhardt’s colleague to be clear about his priorities. Is he, for example, concerned primarily about increasing current output or about redistributing current output? Either might be a worthy goal, but we can’t have a useful debate with someone who won’t tell us what his goals are.
Usually, when economists take policy stands, they start with an efficiency analysis precisly in order to clarify their goals and so make it easier for opponents to identify the locus of their disagreement. They say things like “I support this policy because it’s efficient” or “This policy is inefficient — I estimate the deadweight loss at $X — but I think that much deadweight loss is worth tolerating for the following reasons.” That’s called intellectual honesty. I think it’s a good thing.
Since you mentioned the case of transferring resources to sick poor people, and how the politician might support a policy even if it’s inefficient, because there’s no other way to know who the sick poor people are so you can give them money —
Here’s another reason to support inefficient transfers of resources to sick poor people, even if you happen to know exactly who the sick poor people are.
Suppose you think the poor deserve $10,000 worth of assistance, but that voters in a democracy would only vote to give them $5,000 worth in pure cash. But suppose voters *would* vote to provide $7,000 worth of medical care to the poor (even if it costs, say, $12,000 to provide), and on top of that, voters would support giving the poor $3,000 in cash assistance. Then the only policy sustainable in that democracy, that will give the poor the level of support you think they deserve, would be to provide them with medical care plus cash, instead of just cash.
Either might be a worthy goal, but we can’t have a useful debate with someone who won’t tell us what his goals are.
Words to live by. Thanks for this gem!
Prof. Landsburg,
How do you respond to critics of the efficiency criterion who point out that comparing willingness and ability to pay between rich and poor people for say, food, is clearly unfair. I think a pretty good argument is that if you let prices allocate resources, poor people tend to be better off, and then we could have some redistribution. I was just wondering if you could improve on that argument.
Regarding intellectual honesty, I simply cannot read Reinhardt’s article and each of the linked PDFs from the articles (including an assignment sent to his students) and see it. It is perfectly sensible to write critical pieces on the efficiency concept of economics. But I find it a strain to read such pieces when the author dresses it up as a (plausible) criticism of the “value-free-ness” of the welfare criteria and at the same time writes the criticism as if he himself were being value-free.
I don’t believe personally that one must be value-free so long as they are honest in their presentation. But masquerading as value-free while criticizing implicitly value-laden economic welfare analysis without really recognizing the points you make in your text, or other considerations, is a bit like being the snake-oil salesman that he criticizes economists of being.
Finally, of course the efficiency criteria is an ethical criteria. But some of us do not “bury” this fact as he argues. In fact, it would be nice if Reinhardt laid out for his students (who presumably are not exposed otherwise to the idea) why such an ethical criteria is desirable … or better yet, even if it is viewed as “undesirable” why other criteria (which in these pieces he conveniently leaves out under the guise of being value free) can be more desirable. Simply raising questions about the soundness of the efficiency criteria does not place one on any moral high ground.
The efficiency argument is one of the reasons I’m rather attracted to the Georgist idea of land value taxation (at least, compared to taxes on consumption and income). Land is essentially fixed in quantity, which means there is no deadweight loss for taxing it. Plus, it’s really, *really* hard to hide beachfront property in a Swiss bank account. (Might as well include a Wiki link: http://en.wikipedia.org/wiki/Land_value_taxation )
@ AC: you might start by asking whether economic outcomes are supposed to be “fair” or whether that is desirable at all. That’s for a longer dicussion. A fruitful way to address your point would be to consider what kinds of “fairness” is implied by the pursuit of efficiency and allocating resources by willingness to pay. Certainly such policies are fair producers and “taxpayers” who presumably get to decide what lifestyles to choose and get to keep the fruits of their labor. But it is also useful to understand that the pursuit of efficiency generally leads to lower prices, a wider abundance of food and other goods, and an every improving standard of living for all citizens. So in a static world, the argument you cite above would seem to have a lot more teeth than in the dynamic ever changing world that we actually live in.
There’s much more to say of course, including a discussion of if ability to pay is not to be used, what other alternatives might be worth considering? And should ability to pay be considered for any goods? Why food and health care should be different than other things is not clear to many people?
Finally, a dose of Hayek is in order. How would anyone know what the “true values” of the rich and poor for food be outside of willingness to pay? And do not the behavioralists caution us that people just saying they value something does not necessarily reflect how much they really value it? Furthermore, suppose we agree that we can overcome those two problems, wouldn’t we need to deal with them millions of times a day, every day, forever? These are not static, one-time problems to be solved and then all of us sit back and enjoy some mythical equilibrium. As the world is ever changing and knowledge is dispersed among millions of people, and incomes change regularly and desires change regularly and how people respond to all of those changes, changes regularly, and expectations about all of that data changes regularly, it leads us to understand that we have a problem incredibly more complex than the one you raise or that Reinhardt lays out.
One more thing, such a discussion conveniently leaves out how “income” is obtained in the first place. There is not some fixed amount out there to be distributed to some lucky winners and losers. It is itself produced, and with only small exceptions, everyone has the ability to produce more by serving the interests of others. The “unable to pay” poor people are a far smaller slice of society than is usually presented by capitalist critics.
And of course, all of this ignores any practical and public choice issues in dealing with “solutions” to the supposed unfairness of the situation you cite.
Thank you for making this post, Steve_Landsburg, and I wish more economists would mention a caveat like this when justifying something on the basis of relative efficiency.
In most cases, what bothers me about a policy is not so much that I disagree with the goals, but that it pursues them extremely wastefully. (“If you want to help poor, beleaguered group X, fine, but why do you also have to impose avoidable costs on everyone else?”)
Still, It seems to be contradictory to say that an inefficient policy can be improved on without making anyone worse off … and then go on to list ways in which people become worse off through ideal improvements. Paying off noisemakers isn’t so efficient when you consider the long-term costs of reward extortion, and this should be part of the efficiency calculation, not something you “worry about” once you know the result.
I have a problem with how honest and quick to respond your fictional politician is…
I like this post. I agree Reinhardt’s post is largely a straw man argument. But in addition it suffers from what I call the pundit’s one handed gambit, which is to identify flaws only in the analyses one disaagrees with, leave your own preference unexamined and imply it is free from the same defects. But of course, a claim about the distributional effects of any proposed policy in a nation of 300 million is as fraught with uncertainty and arbitrariness of definitions and categorization as a claim in favor of efficiency.
I do not understand what alternative approach Reinhardt is suggesting. Suppose we do a “constructivist” study of a policy that finds group A loses $100K and group B gains $10K. Then what? We shut up? (A defensible view, but it would require that ALL economists shut up on policy matters, including Paul Krugman.) We suggest distributional weights? (See how far that gets you.) Or we say, correctly IMO, that it will cost society as a whole $90K to do this policy?
Why do we have so many Republicans who hate “spending” but love “tax cuts”?
We could pay every homeowner a certain amount of money. That would be bad because the spending would add to the deficit. We could give every homeowner a deduction for mortgage interest. That would not be bad because it is not spending.
Obviously, we are encouraging people to take out a bigger mortgage than if we didn’t have a mortgage interest deduction.
I used to make my living by transferring tax benefits from people who couldn’t use them to people who could. It was a very inefficient way to do things but it sure paid the middleman well.
There are so many places where paying someone cash would be better than messing with the tax code. I am not making a moral judgement on whether those tax breaks are good or bad. I am just saying they are more efficient that messing with the tax code.
Neil Wilson: Giving every homeowner a certain amount of money is not an example of government spending. It’s an example of a transfer payment. Government spending takes place only when the government consumes resources.
You are right that the fiscal consequences of transfer payments are of the same type as the fiscal consequences of tax cuts; a transfer payment *is* a negative tax. But the consequences of actual government spending are far different.
Using efficiency as a criterion for public policy analysis is incoherent. Pareto or Kaldor-Hicks efficiency is achieved in equilibrium with rational utility-maximizing agents. That is, the assumption of rational utility-maximizing agents immediately implies that whatever is observed is efficient. Thus, on the assumption of rational utility-maximizing agents, anything observed must be efficient, so no “efficiency analysis” can ever recommend a change to the status quo. If some move away from the status quo would be an improvement in efficiency, then the agents in the model aren’t rational utility-maximizers, in which case efficiency has no meaning anyway.
One might try to get around this by modeling the economist as an agent that aides the other people in the model in searching for the best next set of transactions, but this also leads to incoherence. If the economist’s recommendations are rejected, for instance, then he is by definition wrong (on the assumption of rational utility-maximizing agents). Are you, Mr. Landsburg, willing to admit that all of the policy change you’ve recommended that have been rejected—have been wrong?
I think the basic problem is that people confuse inefficiency with stupidity. People think that just because policies are stupid, than they must be inefficient in the economic sense, but this isn’t the case at all. Stupidity is just another constraint: when rational agents maximize utility, given how stupid they are (which might be really, really stupid), then the outcome is still efficient, although often quite stupid.
Rapscallion:
That is, the assumption of rational utility-maximizing agents immediately implies that whatever is observed is efficient.
This is certainly false. Try Googling “Prisoner’s Dilemma” for the simplest counterexample.
Thanks for this post Steve. This, and Reinhardt’s post, clarifies a lot.
Dr. Landsburg wrote:
“‘That is, the assumption of rational utility-maximizing agents immediately implies that whatever is observed is efficient’.
This is certainly false. Try Googling “Prisoner’s Dilemma” for the simplest counterexample.”
This is only under a different and non-relevant definition of efficiency.
The “efficient” solution to the prisoner’s dilemma is out-of-equilibrium, and hence impossible. If there is a way out of the prisoner’s dilemma, then it is not a prisoner’s dilemma.
Comparing different states that are attainable under different sets of constraints is comparing apples and oranges. What’s relevant are the attainable states under the actually-existing constraints.
Umbeck and Staten term the two types of inefficiency N-inefficiency (inefficient relative to equilibrium with different constraints) and P-inefficiency (inefficient in the same model). P-inefficiency can never be observed, and N-inefficiency is irrelevant because it assumes a different world.
Economic Inefficiency: A Failure of Economists
Michael Staten and John Umbeck
The Journal of Economic Education
Vol. 20, No. 1 (Winter, 1989), pp. 57-72
Prof. Landsburg, are you able to justify the following tenet of financial economics that has been bugging me ever since I first started studying the subject?
Notwithstanding the disincentives that could get in the way of this goal, I have no idea how this fosters a result that is in line with the efficiency criterion (or any set of economic principles). And yet it is more or less taken as a given and forms the basis for every idea derived therein.
I have a hard time believing this principle would necessarily obtain an efficient result, let alone a desirable one. One can easily think of stakeholders of a firm’s decisions who don’t happen to own stock; managers, employees, customers, suppliers, etc. So why the fixation on shareholders?
I have yet to find a satisfying answer.
I am not sure why Steve Landsburg wrote this post, because I do not believe that we actually disagree on fundamentals.
In the numerical example he offers, he identifies who wins and loses and by how much, and he notes that it is useful for economists to perform these exercises, in the hope that on the basis of those estimates they might devise a private or public compensatory scheme that would convert a policy promising a mere potential Pareto improvement into an actual Pareto improvement.
I would certainly not disagree with that approach, because here he proposes that compensation actually be paid — that a mere potential Pareto improvement be converted into an actual Pareto improvement. The Kaldor-Hicks problem then does not even arise. But how many public policies actually are structured that felicitous way these days?
Indeed, in my post I noted:
“Strict constructionists argue that their analyses should confine themselves strictly to positive (that is, descriptive) analysis: identify who wins and who loses from a public policy, and how much, but leave judgments about the social merits of the policy to politicians.”
The question can be raised why we even need a criterion such as “efficiency,” after having identified the winners and losers from a proposed policy, and the size of the gains and losses, especially when that word “efficiency” is applied to mere potential Pareto improvements?
As Ken Arrow warns us in his 1963 essay on health care, “A definition is just a definition, but when the definiendum is a word already in common use with favorable connotations, it is clear that we are really trying to be persuasive; we are implicitly recommending the achievement of optimal states.”
The distributional implication of a policy yielding a mere potential Pareto improvement may be such that a majority of voters would not view the policy as desirable. So why then label it with a word that, in Arrow’s words, “is already in common use with a highly favorable connotation”? Why not simply describe the distributional implications of the policy and leave it at that? By draping the policy in the flag of efficiency, are we not trying to persuade people to adopt it? Should we be doing that?
With respect to his example of subsidizing end-of-life care for poor people (or health care for poor people in general), I think commentator Bennett Haselton is right on the mark. Sweet and kind as we economists try to be, we dream of policies that would maximize the utility accruing to the poor per dollar of taxes spent on or transferred to them. I doubt that the majority of tax-paying voters in any democracy would posit that same maximand for their tax dollars. Just imagine if the American Economic Association proposed to Congress and the public that the huge Medicare and Medicaid budgets be paid out to the curent beneficiaries of these programs in the form of cash thart could be spent on anything. The very idea would be DOA.
We economists are arrogant to think that when politicians prefer benefits in kind to cash payments, they have not thought deeply about these issues and need our clever analyses to show them the error of their ways. Chances are that politicians understand the preferences of voters better than do economists.
@ rapscallion: I think you’d enjoy reading Oskar Morgenstern’s criticisms of the idea of static equilibrium in the 1930s.
Morgenstern argued quite that perfect foresight and movements toward equilibrium were quite incompatible. If you (agents) have perfect foresight, then all know what you will do and vice versa, so that we either have permanent regress or permanent equilibrium … thus the process of moving to equilibrium is rendered superfluous.
Hayek, of course, shed much light on the nature of this criticism, and the idea that knowledge is both subjective and dispersed can be synthesized to understand that “correct foresight” is not a PRE-condition which must exist in order that equilibrium may be achieved, but rather it is the defining characteristic of a state of equilibrium.
And given that the world is constantly in flux, coupled with this dispersed and subjective knowledge, it is hard to argue that we are ever in such a mythical state of static equilibrium.
Wintercow,
Thanks for the recommendation. I’ll look at Morgenstern if I get a chance.
I do think, though, that the basic logic of the critique of efficiency analysis stands no matter how fancy you make the utility-maximizing models (e.g. adding uncertainty, time). I still think it’s a non sequitur to argue that:
According to my economic model, which assumes rational utility-maximizing agents, people will choose policy Y, which is different from the observed status-quo policy X. Therefore, we should change to policy Y.
The problem is that the fact that people are currently NOT choosing policy Y MUST mean that either:
1) People are not rational utility-maximizers.
2) The ancillary premises relating to the nature of the decision problems they face in your model are incorrect.
If 1 is the case, then efficiency has no substantive application in the real world. If 2 is the case, you have to go back to the drawing board and get a different model. It doesn’t matter if the policy incorporates time and uncertainty, so that you are just recommending a different policy function based on ex-ante expectations: if people are doing the best they can, then they are doing the best they can, and hence their decisions can’t be improved.
Ultimately, I think efficiency analysis is an attempt to derive “ought”’s from “is”’s. If economics is a science, then the purpose of utility functions is to predict and describe how people behave. The best utility functions are those which come closest to predicting observed behavior. When people behave differently from what’s predicted, the model is wrong, not the people.
Imagine how you’d react to a primatologist arguing that apes would do better for themselves if they learned how to make air-conditioners. You’d probably object that though that may be true, apes are too stupid to make air-conditioners, so it’s an irrelevant point. Similarly, it’s also irrelevant if people would be better off if there were different constraints on behavior that allowed for improved equilibria. The point is that the constraints we have are the one’s we’re stuck with.
Inefficiency, understood in the proper sense, can never be observed.
Uwe Reinhardt:
Thanks very much for this. A few quick comments:
I would certainly not disagree with that approach, because here he proposes that compensation actually be paid — that a mere potential Pareto improvement be converted into an actual Pareto improvement. The Kaldor-Hicks problem then does not even arise. But how many public policies actually are structured that felicitous way these days?
I am proposing a bit more than you acknowledge. In the Bill Gates example, I am proposing that efficiency-based reasoning might well lead us to the following useful sequence of thoughts:
1) Let’s make Bill turn off his music!
2) Wait—that’s inefficient. Maybe it would be better to let him play his music and pay compensation.
3) Hmmm….somehow taking money from Bill and giving it to his neighbors doesn’t feel right to us. (We won’t *necessarily* say this, but we *might*, and I want to consider the case where we do. Maybe we believe that the burden of compensating the neighbors should be spread more widely, or that the money we take from Bill could be put to better use, or that transfers of this magnitude just don’t sit well with us.)
4) We are now forced to recognize that our original command-and-control anti-music policy is certainly inferior to a compensation scheme that sits poorly with us. Maybe that means we shouldn’t feel so good about that command-and-control policy after all.
The fact that the compensation is never paid does not change the fact that *contemplating* the compensation forces us to think hard about what we really want to accomplish and whether we’re really willing to pay the price of accomplishing it.
So I think it’s important for us to keep on contemplating these compensation schemes, even when they have no chance of being implemented. That, of course, is exactly what efficiency analysis is all about.
Rapscallion:
Inefficiency, understood in the proper sense, can never be observed.
You are using the word “inefficiency” in a non-standard way. I think I understand how you’re using it, and I think I understand why you believe this is the right way to think about it. But then you are led to this:
Inefficiency, understood in the proper sense, can never be observed.
which makes your notion of inefficiency substantially less interesting than the standard one.
To Steve Landsburg:
You set up a case in which a noise-making Bill Gates could, in principle, pay those who suffer from his noise an amount of money that would make them put up with the noise.
But then you argue that taking money from Bill and giving it to his neighbors doesn’t feel right to us. Who is the “us” here? And whoever the “us” are, who entitled them to make that call? What about the views of the neighbors?
Then you go on: “Maybe we believe that the burden of compensating the neighbors should be spread more widely, or that the money we take from Bill could be put to better use, or that transfers of this magnitude just don’t sit well with us.”
Now, by “spreading the burden of compensating the neighbors,” do you mean general tax payers should compensate them, rather than Bill Gates?
And by “putting the money we take from Bill to better use,” who entitled “us” (whoever we are) to say these other uses are “better?” Again, who is the “us”?
Finally, when you say ” … or transfers of this magnitude [I guess taking the money from Bill] just don’t sit well with us.” Do you mean that when all is said and done Bill Gates just should be allowed to blare noise at his neighbors?
Finally, you write: “We are now forced to recognize that our original command-and-control anti-music policy is certainly inferior to a compensation scheme that sits poorly with us. Maybe that means we shouldn’t feel so good about that command-and-control policy after all.”
Do you believe Bill Gates’ neighbors will share that judgment — and a pure judgment it is?
Do you really think society needs economists to make all those judgments? If we economists informed society (e.g., the relevant town council)of our estimate, made as economists, of what Bill Gates would be willing to pay to be allowed to make the noise and what bribe his neighbors would require to allow him to do so, why is that not enough? Why do we have to use ther word “efficiency” at all to accomplish that? And if “us” is represented by the town council, why not let it make the call what to do in this instance?
That’s really all the strict constructionists argue for.They do not argue that economists should refrain from estimating the benefits and costs of public policies and the incidence of both. Far from it. They merely argue that it is not up to economists to make the judgment call whether one policy is “better” than an alternative, with appeal to sopmething Kaldor-Hicks call “efficient.” That’s all.
Uwe Reinhardt:
But then you argue that taking money from Bill and giving it to his neighbors doesn’t feel right to us. Who is the “us” here?
The “us” is whoever is making this policy decision. I am suggesting that it’s the economist’s job to force the policymaker to think harder about what he’s trying to accomplish, think about whether there might be a better way to accomplish it, and be honest with the voters about what he’s aiming for. And I am suggesting that efficiency analysis is a very effective way to do all that.
Steve,
Thanks for responding. I just want to make a few more points:
i) I’m not using a weird definition of efficiency; I’m just pointing out that economists often switch between different definitions in misleading ways. In Umbek and Staten’s terminiology (see the citation in my post above), P-inefficiency is plain old Pareto/Kaldor-Hick’s inefficiency whereas N-inefficiency is the inefficiency that makes second-best worlds inferior to first-best worlds. For instance, the argument that an observed monopoly is inefficient is wrong because it uses N-inefficiency as the relevant criterion when P-inefficiency implies that the observed monopoly is the best that can be done given the actually-existing constraints. If the costs outweighed the benefits, agents could pay the monopolist to increase production or abandon the monopolistic power. If you live in a world where this doesn’t happen, then the transaction costs of the exchange must make this exchange INEFFICIENT—so monopoly is the efficient outcome. Pointing out that in some other world where costs were different, the monopoly would go away is beside the point.
ii) I disagree that the logic of this analysis make efficiency analysis uninteresting or unfruitful. Rather, it clarifies what the proper methodology of economic analysis should be: Assume that people are rational utility-maximizers and figure out why what one observes is efficient on that basis. For instance, if we observe that Bill Gates is not paid to turn down the music, we may surmise that it is inefficient for him to be paid to do so. Speculating as to why this might be will lead to richer models that might generate empirical predictions. For instance, perhaps Bill Gates’s neighbor’s are not likely to live near Bill long enough for the payment to be worth it. This might lead one to speculate that people who are more likely to be domiciled in the same place for the long-term are more likely to make these sorts of transactions with their neighbors, which suggests testable empirical hypotheses.
iii) I really think the logic is pretty ineluctable. Name something about the world that you think is Pareto or Kaldor-Hicks Inefficient. There must be some welfare-improving trades that are not being made; you must think that money is being left on the table. Why aren’t they being made? Are people being fundamentally irrational in not satisfying their utility functions? If they aren’t satisfying their supposed utility-functions, why don’t we posit some new utility functions? How can one rely on observed price data to argue for inefficiency, when the very idea of observed inefficiency supposes that sometimes prices don’t reflect true willingness-to-pay?
I do not in any way disagree with you that helping decision makers think hard about and articulate the goals they wish to pursue and, once they have articulated those goals, to help them reach those goals in a way that minimizes the use of real resources getting there. Nothing I wrote suggests otherwise.
On the other hand, telling them (or students, future decision makers) that:
“When Jack gains $10 and Jill loses $5, social gains increase by $5, so the policy is a good one. When Jack gains $10 and Jill loses $15, there is a deadweight loss of $5, so the policy is bad”
is not exactly the way I would go about my work. That is where you and I differ, fundamentally.
I would prefer to tell decision makers that “if you adopt this policy, it will make Jack gain $10 and make Jill lose $5.” I would then let decision makers decide whether or not that is a good thing. I think neither God nor my graduate school appointed me to that position. In fact my professor, the late James Tobin, taught me that I am not entitled to make those calls.
> efficiency analysis is a very effective way to do all that
Yeah, but Uwe’s saying, “Why not just call it cost-benefit analysis”?
But the deeper problem is the failure, in practice, of most economists to include any utility function at all in their efficiency analyses. Professor Landsburg makes analyses of economic efficiency look oh so reasonable by including one, and one that flattens out on top, at that. (Though natural log would probably do a far better job than log of representing the profound differences between rich and poor people’s effective utility functions.)
In fact, the huge preponderance of economic research that powerful policy-makers are exposed to has never been within spear-chucking distance of a utility function. It generally assumes that every dollar buys one util — that a second or third Lamborghini has the same utility as feeding your family for a decade or two, or putting your kids through college.
It must be true, right? The efficient market will force it to be true, eventually…
And the economics profession is decidedly at fault for that absence. The word “utility” doesn’t even appear in Mankiw’s textbook, for instance, until page 442. And then its impact on welfare economics is largely relegated to a sidebar on page 465, one that silently assumes Kaldor-Hicks.
It’s like a physics textbook asserting that a bowling ball falls faster than a feather because it’s heavier, and then mentioning friction, in passing, on page 400.
Uwe Reinhardt: Your quote (from my textbook) illustrates the definition of the efficiency criterion. The efficiency criterion declares the policy in question a bad one. If students are going to understand the efficiency criterion, they need to understand such examples.
They also need to understand that the efficiency criterion is only one of many normative criteria, that those criteria sometimes disagree about what’s good and what’s bad, and that in practice no economist is 100% committed to any one of these criteria.
But to understand this subject, they need (at a minimum) to be able to figure out what each criterion declares good or bad in (at least) some simple stylized examples. I stand by the stylized example of Jack and Jill as a clear and useful illustration of the efficiency criterion.
Asymptosis: Utility functions have nothing to do with the efficiency criterion.
As for this:
the huge preponderance of economic research … generally assumes that every dollar buys one util
I have never in my life seen an example of an economic research paper that makes this assumption. Perhaps you could provide an example.
I’d like to borrow an argument from David Friedman’s Price Theory textbook (I believe it’s also in his “Hidden Order,” and I hope he’d agree that this is an accurate summary.) We ought to make as many efficient moves as possible, because it’s likely that the biases of individual moves will wash out for some particular person (they will average to zero). If we scrap a thousand public policies, some may have a net cost to you, some a net benefit, but there is unlikely to be a large bias for any particular person. You likely lose a few big friends but thousands of tiny enemies, to at least a small net benefit.
On the other hand, if we give too much deference to politicians and “their knowledge of their constituents’ interests,” they are likely to choose a lot of inefficient moves that happen to benefit their own constituents, and the people who pay the costs are unlikely to be aware of these costs or to have sufficient incentive to DO something about it (on the off chance that they do know the costs).
I don’t buy the argument (Kenneth Arrow’s apparently?) that ‘efficiency’ has a positive connotation. It actually sounds very cold and autocratic. If you tell someone that you’re dealing with them as “efficiently” as possible, they’re unlikely to thank you. Anyway, something can seem cold and unfeeling and still be the correct criterion.
It would be hard to argue that “efficiency” had negative connotations! In this context, efficiency is definately a loaded term.
There are two problems. Problem 1 is the measurement problem: we cannot asses definatively all the costs and benefits. This introduces a bias into the determination of efficiency, which I think Reinhardt discusses with in the “strict constructionist” school. It is impossible to arrive at the efficiency calculation without some value judgements, so the term is not inherently neutral. That does not necessarily mean that we should not make the attempt.
Problem 2 is the idea that the most efficient result is the “best” result. There seems to be general agreement that this is not necessarliy the case, and other coinsiderations may be important. For example, general distaste at large payments from Bill Gates. However, is this not really the same as problem 1? All you need to do is include these factors into the efficiency calculation.
It seems by definition that the most efficient result is the “best” result, if you have included all the costs and benefits. But we have no way to arrive at some of the costs and benefits, so we cannot do the full efficiency calculation without huge assumptions and introduction of bias. So if we use the efficiency calculation for policy, we must be aware that it is only a partial answer.
The third problem arises when politics comes into the game. It is difficult to argue for a less than efficient solution. If the results of the calculation are labelled “the most efficient”, it gives them great weight. All it is really saying is this is the most efficient, taking into consideration those things we can easily (or possible with difficulty) measure. In other words, the partial nature of the estimate should be made clear. Saying “the most efficient” does not do this, and creates the impression of false accuracy (or possibly precision).
Following Reinhardt’s logic, economists should also stop doing national income accounting, because it adds together increases in dollar outputs by some and losses in outputs by others and ignores who gains and who loses. We should shut up about countercyclical policy as well, becauses it is about the levels and growth rates of such aggregates. All we should do is calculate an enormous list of who gains and who loses each year, or from each possible fiscal and monetary policy, and shut up. For example, the Fed’s low interest rate policy is hurting savers and helping spenders. We should stop there. We have no business saying that it is good or bad for the economy.
Response to Steve Landsburg:
Once again, I do not at all disagree that it is important to spell out tyo students what the Kaldor-Hicks efficiency criterion implies, which is precisely why I quoted you, because you make it quite clear.
I sometimes use a different example, arguing that if you would require $500 in compensation to let me hit you in the gut and I would be willing to pay $800 for the privilege of doing so, then hitting you and walking away without paying you has increased social welfare in America by $300. To be kind, though, I might give you a copy of Kaldor’s original paper, to help you understand what had just happened.
The problem is that economists do not invariably spell out these stark implications when they use Kaldor-Hicks efficuiency in their applied work.
Professor Landsburg:
I would just ask how many of those papers specify any utility function, as you do (appropriately) even in your toy model. How many even mention utility?
http://www.google.com/search?&as_q=&as_epq=progressive+taxation&as_sitesearch=econpapers.repec.org#num=100&hl=en&q=progressive-taxation+site%3Aeconpapers.repec.org
913 results
http://www.google.com/search?q=progressive-taxation+%2Butility+site%3Aeconpapers.repec.org
85 results
While you may assume that they all assume variable utility, how many include utility functions in their calculations of costs and benefits?
And to quote Professor Reinhardt in his takedown of textbook methodology:
“To assume that the deadweight loss C expressed in money terms also reflects a net social loss in human happiness (utilities) *tacitly assumes that* … the marginal utility of money wealth is constant across income classes.”
Emphasis mine.
To Asymptosis:
I side with you in your assertion that economists implicitly make some strange assumptions about utility when they blithely integrate under the market demand curve and call the area under it (up to some volume Q) the “social benefit” produced by that volume, where by “social benefit” they clearly mean a measure of human well being.
I tell my students to think of a point on the market demand curve as the maximum bid price some human being with a name who is represented by that point is willing to pay for a unit of the thing being traded. Although one human being may appear at several points on the curve, it is easiest to assume that every person bids on and possibly buys but one unit per period so that every point on the curve represents a different human being. I also remind my students that wealth and, thus, ability to offer high bid prices varies among people represented by the curve.
It is uncontroversial to call the area under the demand curve (up to Q) the maximum revenue that could be extracted from consumers under perfect price discrimination. Call it maximum extractable revenue (MER).
It is quite another thing, though, so impute to that area a measure of social welfare, of collective human well being, as is our profession’s wont. Look at any of the text books using the little triangles to explain the social benefits of free trade, for example. We blithely assume that the areas under the demand curves do measure social welfare, do we not? The whole case for free trade we make rests on that assumption.
If Jones bids $100 for the thing being traded and Chen only $50, what does that really tell us about the human well being behind these numbers? It merely tells us that for whatever reason, Jones is both able and willing to surrender twice as much money for the thing than Chen is able and willing to surrender. But it tells us nothing about how much each of them value the thing in the sense that they crave or appreciate the thing – the “utility” it yields them. For all we know, a poor Mr. Chen may derive much more happiness from having the thing than a wealthy Mr. Jones.
We would be on safer grounds if we could assume that a dollar represented the same intensity of pleasure (utility) for everyone participating in this market – for Donald Trump as well as a waitress. In other words, we would assume that whatever The Donald’s and the waitress’ wealth may be, an additional dollar of wealth would yield the same degree of pleasure to each, so that when The Donald bids 10 times as much for a thing than does the waitress, we can safely assume that he derives 10 times the intensity of pleasure from the thing.
If we could assume all that, we could then add up horizontally the bid prices represented by a demand curve (up to volume Q) and pretend that the sum is a dollar measure of something we might call “collective pleasure” or “social welfare” and we could do the little triangle shows we put on in Econ 101 with a straight face. That much would be purely mechanical, even if heroic.
Ethical doctrine enters, however, when we view the dollar value of that construct as something to be maximized through public policy and our profession to be licensed to make sure that the economy is run that way. That assumption is anything but controversial.
Asymptosis: I have no idea what you mean by “include utility functions in their calculations of costs and benefits”. Costs and benefits are measured in dollars, not in utils.
Asymptosis:
“To assume that the deadweight loss C expressed in money terms also reflects a net social loss in human happiness (utilities) *tacitly assumes that* … the marginal utility of money wealth is constant across income classes.”
Right. Which is why absolutely nobody assumes that the deadweigh loss expressed in money terms also reflects a net social loss in human happiness. You are attacking a straw man.
Did you miss this passage from the post you’re replying to?
no economist thinks those numbers in any way measure Bill’s subjective enjoyment of his music or your subjective annoyance. Only a crazy person would think such a thing, and I’ve never met anybody who’s that crazy in that particular way.
Uwe Reinhardt:
economists implicitly make some strange assumptions about utility when they blithely integrate under the market demand curve and call the area under it (up to some volume Q) the “social benefit” produced by that volume, where by “social benefit” they clearly mean a measure of human well being.
No, actually by “social benefit” they mean “social benefit”, which is a well-defined term of art and nothing so vague or nebulous as a measure of human well being.
It is quite another thing, though, so impute to that area a measure of social welfare, of collective human well being, as is our profession’s wont.
It is a measure of social welfare by definition. It is not a measure of collective human well being first because I have no idea what that even *means* and second because if it *does* mean anything, it’s certainly not the same thing as this. What you call “our profession’s wont” I call a bit of silliness that I’ve never heard from anyone.
The whole case for free trade rests on that assumption.
No, actually it doesn’t.
If Jones bids $100 for the thing being traded and Chen only $50, what does that really tell us about the human well being behind these numbers? It merely tells us that for whatever reason, Jones is both able and willing to surrender twice as much money for the thing than Chen is able and willing to surrender.
No, it tells us quite a bit more. It guides us to consider alternative policies that could make both Jones and Chen better off. It also guides us to compare those policies to still other policies we might not otherwise have considered, or considered in quite the same light.
We would be on safer grounds if we could assume that a dollar represented the same intensity of pleasure (utility) for everyone participating in this market
Wrong on two counts. First you are conflating utility with “intensity of pleasure” (whatever that means) when nobody in the world pretends that it represents any such thing. A utility function is a largely arbitrary assignment of cardinal values to ordinal preferences. Period.
In other words, we would assume that whatever The Donald’s and the waitress’ wealth may be, an additional dollar of wealth would yield the same degree of pleasure to each, so that when The Donald bids 10 times as much for a thing than does the waitress, we can safely assume that he derives 10 times the intensity of pleasure from the thing.
As my Bill Gates example emphasizes, this is entirely off the point. Nobody assumes this, nobody needs to assume this, nobody (as far as I know) even knows what it would *mean* to assume it.
something we might call “collective pleasure” or “social welfare”
“Social welfare” already has a perfectly well defined meaning. Why would you want to use it to mean something different entirely, like “collective pleasure”?
Neil: Bravo.
To Steve Landsberg:
You argue that by “social benefit” economists meand just that “social benefit,” which is a well defined term. Just for the record, could you define it for us?
The same goes for “social welfare.” You say it has a well defined meaning. Could you give us that one as well?
It might clear up a thing or two.
We saw this recently play out over the health insurance debate.
Using Landsburg’s analytical method, one attempts to boil down legislation to its true nature or core set of goals, and then debate those goals. For instance, the debate may have focused on whether it was a good idea to have the federal govt redistribute health care from those who have it to the poor.
Using Reinhardt’s logic, you never attempt to question the true goal of a bill, thus allowing politicians to use their feigned goals and attach a number next to it (as if it were a given the legislation will accomplish even its feigned goals). This allows politicians to say things like “We will improve healthcare for the poor *for only $900b*”.
The latter leads to a debate about the role of govt, property and natural rights, redistribution, etc. The former leads to lots of ad-hominem attacks since only evil people are against improving healthcare for the poor.
To Neil:
When you write “We have no business saying that it is good or bad for the economy,” you seem to treat the economy as a person. Please define for me what is “the economy”?
To argue that economists should be careful and transparent in their normative analyses is not to argue that they should not engage in cost benefit analysis.
Frankly, I find statements such as “this is good for the economy” rather vacuuous. We can do better than that.
We might explain how a low-interest rate policy might kindle economic activity and employment, but also point out that it has the effect of hurting savers (or insurance companies) and benefiting the shareholders and employees of banks. I find that more informative than merely proclaiming “lower interest rates because it is good for the economy.”
Uwe Reinhardt:
You argue that by “social benefit” economists meand just that “social benefit,” which is a well defined term. Just for the record, could you define it for us?
The same goes for “social welfare.” You say it has a well defined meaning. Could you give us that one as well?
Let A be a non-status-quo policy and let John be an individual with wealth W. John’s willingness-to-pay for policy A is that amount x such that John is indifferent between living in the status quo world on the one hand, and living in a world with policy A and whith his wealth reduced to W-x on the other hand. The social benefit of policy A is the sum of individual willingness-to-pay, where the sum is taken over all individuals for whom willingness-to-pay is positive. The social cost of policy A is the sum of individual willingness-to-pay, where the sum is taken over all individuals for whom willingness-to-pay is negative. The “social welfare gain” from adopting policy A is the social benefit minus the social cost. I’m not sure “social welfare” means anything at all, though it can be given an ad hoc meaning by assigning some arbitrary number to the current state of the world. It’s the social *gains* that are meaningful, not that arbitrary constant.
There is another more-or-less (but not exactly) equivalent formulation in terms of willingness-to-accept. When income effects are small, the two formulations provably give results that are approximately equal, and we tend to use them interchangeably (though stopping every now and then to check that our sloppiness does no material harm). When the two formulations differ substantially there is, as far as I am aware, no single generally agreed upon definition of social benefit or social gain, though one can always select one of the small number of choices and use it in an argument, provided one is clear about what choice has been made.
While we’re on the subject of definitions, a “normative criterion” is a rule that assigns the word “good”. “bad” or “neutral” to any prospective policy change. The “efficiency criterion” is that normative criterion that assigns the word “good” to those policy changes for which the social gain is positive and “bad” to those policy changes for which the social gain is negative.
Note that none of this has anything to do with either utility or pleasure (and incidentally that utility is not, and is not supposed to be, a measurement of “pleasure”), except insofar as that, in any model where preferences are represented by a utility function, John’s willingness-to-pay for a policy will be positive if and only if that policy raises his utility. Beyond that, nothing quantitative can (in general) be said, and no economist pretends otherwise.
Uwe Reinhardt: Re your reply to Neil, we are in total agreement about the use of vacuous phrases like “good for the economy”. I think we are also perhaps in agreement about the use of vacuous phrases like “collective pleasure”. The difference is that economists really do use the first formulation, and should be chastised for it, whereas the second formulation, as far as I can tell, is something you just invented.
To Uwe:
My point is that when we treat a dollar as a dollar in measuring efficiency (benefit cost analysis based on the compensation principle), we are not doing anything different than when we treat a dollar as a dollar in national income accounting. The difference is that in efficiency analysis we are adding and subtracting *economic* values rather than commercial values, which is I think a good thing because there are goods and services that have economic value but no commercial value, and sometimes market prices do not fully reflect economic value.
Why do you insist that we must not add economic values but consider only distributions, when we don’t do that for commercial values in national income accounting? I do not think a 300 million element vector measure of economic well-being practical, myself. Aggregating does not prevent us from also considering distribution. I consider efficiency analysis (measuring aggregate economic value) a big improvement over national income accounting myself, and economists should not abandon it.
Steve Landsberg: I concede that I coined the term “collective pleasure.” It isn part of my attempt to understand what might be meant by “social welfare.”
Where does “social” come in here? What does it mean? And what does “welfare” mean in this context?
Is social something different from collective?
I am still looking for your definitions?
Professor Landsburg:
“Costs and benefits are measured in dollars, not in utils.”
That is *exactly* what I am pointing out. (As is, I would presumptuously say, Professor Reinhardt.)
How does this statement square with your previous?:
>>the huge preponderance of economic research … generally assumes that every dollar buys one util
>I have never in my life seen an example of an economic research paper that makes this assumption.
You stipulate that cost/benefit analyses are done in dollars, ignoring utility. But you’ve never seen a paper that (at least silently/implicitly/presumptively) equates those dollars to utility?
I should clarify, by the way: “assumes that every dollar buys one util *or soon will.*” If good X delivers less utility/dollar than good Y, people will stop buying good X until its price drops so its utility/dollar ratio equals that of good Y.
But in the real world, the utility purchased by a rich person buying a second or third Lamborghini is far lower than that of a poor person buying a quality college education for their kid. (I think most everyone would stipulate to this?)
Shouldn’t the price of Lamborghinis plummet?
No. Because rich people and poor people have different utility functions (roughly aggregated in the log function in your toy model).
If a cost-benefit analysis doesn’t (explicitly) include a function like that one, it’s not representing true costs and benefits.
I fully acknowledge the practical difficulties, by the way, of determining accurate utility functions, either individual or aggregate. But we can certainly say that an aggregate function would look something like the log function you use in your toy model — steep on the left, flat on the right.
Uwe Reinhardt:
I am still looking for your definitions?
I gave you definitions of social gain, social cost and social benefit. Welfare means the same thing as social gain (up to an arbitrary additive constant). “Social” is not a term I’m familiar with in this context, except as part of the phrases I’ve already defined.
Aymptosis:
“Costs and benefits are measured in dollars, not in utils.”
That is *exactly* what I am pointing out. (As is, I would presumptuously say, Professor Reinhardt.)
Then you are pointing out something that everybody already knows.
Is it your claim that it’s not *useful* to measure costs and benefits in dollars? If so, you’ve missed the entire point of the exercise, which is that dollars are *transferable* while utils are not. Try reworking my “Bill Gates” example with utils and you’ll see that it doesn’t work, precisely because of the non-transferability.
Asymptosis:
You stipulate that cost/benefit analyses are done in dollars, ignoring utility. But you’ve never seen a paper that (at least silently/implicitly/presumptively) equates those dollars to utility?
Correct. This is about the third time you’ve claimed such papers exist, without providing a single example.
I still don’t really get it. If we have included everything, then the efficient answer must be the best answer. If we were to reject the efficient answer because of other considerations, then surely we should include the value of these considerations within the efficiency calculation?
I presume we don’t because we have no effective way of putting a value on them, so we have to fudge.
Professor Landsburgh:
It’s difficult to cite papers that make only implicit assumptions. My google example was demonstrating that very few papers even mention utility, and far fewer incorporate a utility function like the one that you apparently (and quite rightly) seemed to feel was a necessary part of your toy model.
All the others — the ones that do not transform dollars -> utility in any stated, transparent, systematic way — are implicitly assuming that one dollar equals one util (or will soon — the market will enforce it).
>Is it your claim that it’s not *useful* to measure costs and benefits in dollars?
No, absolutely not. (And: good question.) I would say that, especially in macro, especially in the areas of taxation and redistribution, the failure to include a(n at least aggregate and tentative) utility function in cost/benefit analyses makes those analyses significantly *less* useful, if at least one goal is to estimate and increase sum(people:utility).
We use dollars as a rough measure of utility, just as we use GDP as a rough measure of prosperity. We do that for practical reasons: “prosperity” and “utility” are hard to measure. That should not blind us to the fact that those measures are, arguably, quite systematically distorted by underlying assumptions and resultant practices.
I am pointing out that the implicit equation of dollars to utility in cost-benefit analyses (the failure of most analyses to correct for the utility curve) is one of those systematic distortions — for instance overstating progressive policies’ utility costs to the rich, and understating their utility benefits to the poor, hence potentially understating aggregate benefits.
>If so, you’ve missed the entire point of the exercise, which is that dollars are *transferable* while utils are not. Try reworking my “Bill Gates” example with utils and you’ll see that it doesn’t work, precisely because of the non-transferability.
Assuming that playing loud music and sleeping at night have the same util values for Bill and his neighbor: the value of playing loud music is 1 util. The value of sleeping at night is 2 utils.
Bill’s poor neighbor typically spends $10 per util. Bill spends $10,000.
Should Bill pay the neighbor $20, so the neighbor gets his two utils back — even though it only costs Bill 1/500th of a util? Or should Bill pay the neighbor $20,000? He pays double his normal cost for the one util, and the neighbor gets 1,000 times his normal price for his two utils.
Probably somewhere in between, probably (much) closer to the former than the latter.
The judge or the town council will have to make normative choices based on these and other considerations, but at least they know the parameters. Without those utility calculations, they don’t. (And I would argue that they make these kinds of intuitive utility calculations all the time)
(Don’t get me wrong, by the way: in this extreme example, the first solution creates 1 and 499/500ths of a util. The latter solutions creates 1,998 utils. It shows how a toy thought experiment in micro, especially an extreme one standing by itself, can have ridiculous implications for macro.)
Asymptosis:
Assuming that playing loud music and sleeping at night have the same util values for Bill and his neighbor: the value of playing loud music is 1 util. The value of sleeping at night is 2 utils.
What’s a util?
Asymptosis:
We use dollars as a rough measure of utility
Who’s this “we”? You still haven’t given me a single example of anyone who’s ever done this.
It is, I think, crystal clear that you’ve grossly misinterpreted an entire literature and are determined to go on picking away at the literature of your imagination while ignoring every attempt to explain that nobody’s saying what you think they’re saying. This puts you well along the road to serious crankdom.
Steve Landsburg:
Don’t forget to help this hapless immigrant understand still trying to master the English language what “social” means in “social welfare,” and what “welfare” means, and how the term “social welfare” differs from “collective well being” or “collective pleasure.” I believe, if memory serves me correctly, Bentham expressed his notion of utility as pleasure or pain.
I garbled my previous comment. As they say in Washington “Mistakes were made.”
Here’s what I meant:
Steve Landsburg:
Please don’t forget to help this hapless immigrant (who is still trying to master the English language) understand what “social” means in “social welfare,” and what “welfare” means, and how the term “social welfare” differs from “collective well being” or “collective pleasure.” I believe, if memory serves me correctly, Bentham expressed his notion of utility in terms of “pleasure” and “pain.”
Professor Landsburg:
If you’ll answer one question for me I’ll quit pestering you. (Well, I will anyway, but..)
In your efficiency analysis you felt it necessary to include a utility function.
Is it necessary for others to do likewise in their efficiency analyses?
Uwe Reinhardt: I can no more tell you what “social” means in “social welfare” than I can tell you what “Uw” means in “Uwe”. It’s a single phrase, for which I’ve given you, I think, a clear and precise definition. Who cares what Bentham was thinking?
Asymptosis: Any economic model assumes that people have preferences. A utility function is a convenient way to summarize those preferences. So when you ask whether it’s necessary to include a utility function in an efficiency analysis, the answer is: It’s necessary to include a utility function in *any* economic analysis, whether you’re talking about efficiency or not.
But you are (I am certain) majorly confused about the role of that utility function. Take again the Bill Gates example from my post. I assumed that your night’s sleep is worth $50 to you. I could just as easily have said “Assume that U(W-50,1)=U(W,0) where U=U(wealth, #of nights’ sleep) and W=current wealth.” So the utility function *was right there all along even though I didn’t explicitly mention it*. As it is, I am sure, in all of those 935 papers you tried to show me.
Have you noticed how I answer your questions, but you ignore mine? That’s another one of the ten major warning signs of crankdom.
Several things re: Kremer’s solution
1) That sounds an awful lot like how the mortgage crisis got going.
2) Not sure I buy the cost-benefit analysis. Something’s missing. Take a look at what happened in the mortgage crisis.
3) Perhaps the analysis should be compared to a world without patent protection. Fashion for instance. I know fashion doesn’t require the immense research costs (why do those costs exist?), but I still think we under appreciate what could happen in a world without patent protection.
4) Is the government empowered do this? Is it eminent domain? If not, do you support adding the empowerment through the amendment process or by simply ignoring the Constitution as other encroachments have done in the past?
Seth:
1) I do not understand your analogy with the mortgage crisis.
2) The govt is surely empowered to do this as long as it’s voluntary for the inventors, and probably otherwise also. In any event, who would object to it? It’s wealth-enhancing (in expectation) for pretty much everyone.
Professor Landsberg:
I’ve had similar imputational thoughts regarding your responses and non-reponses, but consciously chose to not state them.
But as promised I’ll now stop pestering you.
All the best to you.
My apologies, Steve. I missed the lengthy definition you gave earlier.
So if we think, for the sake of simplicitly, of a two-person society composed of a very wealthy person who favors policy A and is maximially willing to pay $1,000,000 to see the policy implemented and a poor one who would accept the policy only if paid at least $10,000,then implementing policy A would ipso facto yield a “social welfare gain” of $990,000, even if the waitress were never paid the $10,000.
I wonder how many non-economists would agree. Here I am reminded once again of Ken Arrow’s admonition not to use words that have one meaning to economists and another to lay people.
Uwe Reinhardt:
Here I am reminded once again of Ken Arrow’s admonition not to use words that have one meaning to economists and another to lay people.
In that case, we’ll have to stop use the words “cost”, “price”, “supply”, “demand”, and “equilibrium” at least for starters.
We talk about a lot of precise technical concepts in economics, and we need words for them. The usual procedure is to hijack an English word with a somewhat related meaning and give it a precise definition for use in economics. That’s also the usual procedure in physics and mathematics (the fields I know something about) and, I am guessing, in every other field. Is that the practice you’re objecting to?
Sorry. I posted my previous comment on the wrong thread. Meant to be on the “Idea of the Decade” thread. My mistake.
Mortgage crisis: Quasi governmental organization (Fannie/Freddie) buying mortgages competing with private investors. The difference here is the coin flip. Could work in theory, but I don’t have faith that politics wouldn’t creep in and make things much worse.
I have no issues with the private bidding part. But I wonder which part of the Constitution empowers the government (assuming Federal) to spend taxpayers money on buying out patents.
Steve Landsburg:
So, to round this off,let me ask you this final question:
If, as a result of implementing Policy A, Jack gains $10 and Jill loses $5, does Policy A yield a “social welfare gain” of $5?
Tks.
Uwe
“,then implementing policy A would ipso facto yield a “social welfare gain” of $990,000, even if the waitress were never paid the $10,000.”
Is this correct from an economists viewpoint?
Very inefficient is to not impliment the policy. Waitress as happy as before, but rich man suffers $1,000,000 loss, net loss = $1,000,000
The “efficient” solution is for the wealthy to pay the waitress and implement the policy. Total gain = $990,000. Waitress is as happy as before, rich man also very happy because he gets $1,000,000 benefit for $10,000 cost.
How about implement the policy, but make no payment? Waitress suffers $10,000 loss, rich man gets $1,000,000 benefit. Net gain $990,000. This is the same as in the first case. So presumably the best option is to implement policy A, but don’t bother with the payment and save the transfer costs. This is a position very few would take. Steve says we may not wish to do this because of other factors. What are these, and how do we decide whether to implement the policy?
My original thought was that we must put a value on the “other factors” How would we go about that in this case? The only way I can see is to put a value on each individuals desires independantly of their ability to pay. This is something like the “utils” refered to earlier, but to avoid further confusion, lets call them “quality of life” points, or QP’s. Lets assume that each individuals happiness, or quality of life is equally important, no matter how wealthy they are.
In our 2 person world, we can either implement the policy or not. It might be that the rich person does not care much. $1,000,000 is so trivial that it is hardly worth considering. Lets say he cares 1 QP. The poor person cares deeply. This will affect their life significantly. Say he cares 100 QP’s, or he would lose 100 QP’s if implemented. With only this information, we would surely say “do not implement the policy!” The cost in QP’s to the poor person is greater than the gain to the rich person. Not implementing policy A gives -1 QP, poor person as before, rich person 1QP down.
But we do have more information. We know that the poor person is willing to allow the policy on payment of $10,000. We could say that we are valuing 100 poor persons QPs at $10,000, or $100 each. Rich persons are valued at $1,000,000 each. But crucially , we are also saying that 1 poor persons QP is equal to 1 rich persons QP. Our objective is to maximise total QP.
Armed with the information above, we know that there is an even better solution – have the rich person pay the poor person $10,000 and implement the policy. We know that the benefits to the poor person in QP’s of $10,000 is equal to 100 QP. They cancel out at this point for the poor person. The rich person gains one QP, so this is the best overall with a net gain of 1 QP. It is economically most efficient, and socially most efficient. If the payment is not made, then the total would be -99 QP.
(We have a slight problem here, because if maximising QP is our aim, then we must make the rich man give lots of money to the poor man. If he gives $1,000,000, then he loses 1QP. The poor man gains 10,000 QP. I am starting to see the point of the utility function, and the shape of it, but forget this for now.)
From this example we have seen that following the efficiency calculation has allowed us to arrive at a better solution than without it. We have also seen that efficiency is a means, not an end in itself. I think in policy terms, it is difficult to say this man’s $ is worth less than that man’s $, but this is in effect what we are saying. If in our example it was somehow impossible for the payment to be made, we should follow the very inefficient option of not implementing policy A. The rich man can point and say “but this is inefficient! You get only $10,000 benefit for $1,000,000 cost! How can you choose this?” This is a very powerful argument, which relies on counting that which we can see (the money), and ignoring the unseen (the quality of life).
Uwe: If, as a result of implementing Policy A, Jack gains $10 and Jill loses $5, does Policy A yield a “social welfare gain” of $5?
Yes. This is a direct consequence of the definition and I am extremely puzzled as to why you have to ask.
To Steve Landsburg:
I would like to be able to quote it, Steve. It is a crisp statement of the issue, and you have a penchant for crisp clarifty — also, by the way, on the economic case for the voluntary army. No one has put this as crisply as you have.
Best,
Uwe
I just ran into this hilarious back and forth today and I have to make a post.
I am interested in what Asymptosis had to say about the use of utility functions that have declining marginal utility of money. For example the Ln function. Dollars are a very good at measuring value but there is something to his point about Lamborghinis and a college education.
“In applications, the Kaldor-Hicks criterion and the efficiency criterion amount to the same thing. When Jack gains $10 and Jill loses $5, social gains increase by $5, so the policy is a good one. When Jack gains $10 and Jill loses $15, there is a deadweight loss of $5, so the policy is bad.”
A world can me dreamed up whereby the second situation has a greater social welfare after the transfer than before; where social welfare is defined as the sum of all utility functions, i.e. Jack and Jill’s utility functions.
Let utility (U) be the Ln of the persons wealth in dollars. At time (t) 0 Jack (K) will start off with $1 and Jill $100 (it does not say how much money either had initially so I have taken liberties). At time 1 there will be a transfer of $15 from Jill (J) with $10 going to Jack and $5 will be thrown into the sea where an evil sea creature eats it up (somehow $5 dollars of real wealth is thrown into the sea and we do not care about the monster’s welfare).
t=0
U of K=Ln(1)=0 — U of J=Ln(100)=4.6 —- Social Welfare= 4.6
t=1
U of K=Ln(10)=2.3 — U of J=Ln(85)=4.4 —- Social Welfare= 6.7
There has been a “social gain” and this policy is now good under this reanalysis. It all depends on how you measure what can broadly be called social welfare. Asy’s point is that the first bit of consumption is probably more important to people that the last bit. I recognize it is impossible to really measure this stuff so we have to pay attention to dollar signs. But, there is some important information in this experiment that think think Asy was trying to get at.
Finally, before you go into some tirade on what a utility is, let me say this: I do not know, but you definitely want more of them. I think Asy would say it’s a way to change the dollar so it has decreasing marginal value.
Cornelius:
A world can me dreamed up whereby the second situation has a greater social welfare after the transfer than before; where social welfare is defined as the sum of all utility functions, i.e. Jack and Jill’s utility functions.
You are not dreaming up a different world. All you are dreaming up is a different way to decide between outcomes.
There has been a “social gain” and this policy is now good under this reanalysis.
The fact that a policy is good under your (perfectly reasonable) criterion does not mean that it creates a social gain.
Of course you can define “social gain” any way you want, but since it already has a perfectly good meaning, and since what you’re talking about can be perfectly well described as a “sum of utilities”, it only muddies the water to hijack an existing term and use it in a nonstandard way.
Cornelius: Note, too, that what you and Asymptosis are doing are two very different things. Asymptosis was confused about several things at once, but among others, s/he assumed that there was such a thing as utility that could be measured and added up across people. You, by contrast, are (far more reasonably) simply *defining* utility in a certain way without assuming that it has any prior meaning. (There are much better approaches than yours, but yours is much better than Asymptosis’s.)
But once again—just because you’re doing something reasonable does not mean it’s useful to change the meanings of standard terms. One policy is preferred by an efficiency criterion; a different policy is preferred by your sum-of-utilities criterion. I can easily write down a different sum-of-utilities criterion, with different utility functions, that disagrees with yours. One criterion endorses one outcome; another endorses another. That’s not paradoxical, but it can get confusing if you hijack vocabulary from one criterion to another.