Thursday, December 20, 2012

Munger on Externalities

HT - Kevin Rollins


I used to blog on externalities a lot on here. It's a very important part of economics, and a lot of libertarians have a tendency to treat it (and Pigou) like voodoo. It's very nice to see that this is not true of Mike Munger's new video on externalities. Not only does it outline the theory well, but it also gives Pigou the credit he deserves. He's long been treated like a dummy by people, particularly in the public choice theory tradition, but he made most of the major points that they would only later come to! I think an important factor in this turn may be an article published earlier this year on Pigou by Backhouse and Medema in the Cambridge Journal of Economics.

I would raise two critiques with the video:

1. I strongly disagree with the contention at about 4:20 that economists rarely talk about the trouble governments have in setting the ideal Pigovian tax rate. That just sounds bonkers to me. I don't know how they teach externalities at Duke, but my experience is that this is bullet point number one in discussions of the pros and cons of Pigovian taxes. I really don't know what he could be thinking here.

2. It's only a five minute video, but I'd go more into the limitations of Coase. Coase is often treated as having made Pigovian taxation untenable for serious people. This is wrong, of course. Policy is always about social consensus and our values and ethics. Sure bargaining can solve some of these problems, but sometimes we simply don't think its right that person who bears a cost has to compensate the aggressor for damages done! In some circumstances we find that so distasteful we call it "extortion"! Coase is important for the relevance of transaction costs to other areas of economics, but he definitely didn't toss Pigou out the window. Pigou gets a "yes, but..." and it would have been nice to add at least a sentence or two of "yes, but..." to Coase.

27 comments:

  1. On your first criticism, I don't recall that problem with Piguvian taxation being featured prominently in my econ courses. (Maybe it's more commonly discussed at the graduate level?)

    On your second problem, I think this is largely a function of the length of the video. But on the other hand, this is addressed when Munger talks about manners. One reading of it is that the manners norm assigns the initial property rights (right to less noise rather than right to make noise) and the Coasian bargain comes after that. I'm not sure that's the most natural reading of the video, but that's what came to my mind when watching it.

    ReplyDelete
    Replies
    1. I don't know - the intro micro class I TAed last spring definitely went over it. I thought of mentioning it in the post, but actually I don't know if we explicitly talked about this in grad micro. I remember a lot of "solve for the optimal price" exercises, etc. I do remember it was discussed in the book, but I'm not sure it made it into our discussions because for most economists "the government is not a wizard" is sort of implicit! I know I got the critique at W&M when I was an undergrad.

      I don't know, this is such a basic point I have huge doubts about people that claim nobody talks about it or that it's a unique insight. You don't need some kind of deep esoteric insight to know government doesn't have this info. Maybe I'm being overconfident in calling them overconfident on this point, but I think the logic of externalities itself is a lot trickier and less intuitive than the case that governments don't have this information.

      As for the second problem, ya I think you're right but it wouldn't have taken all that long. Obviously they have different priorities to push too. When half the video is spent saying "yes, but..." to Pigou it gets a little silly to say you can't spare a sentence or two more on Coase. But that's not a biggie. I just like to share my immediate reactions on the blog.

      Delete
    2. The other thing about Coase is just that he is so often presented as the guy that proved Pigou wrong, which is just not correct. And I think there's some value to countering that in the marketplace of ideas. Munger of course didn't go that far in this one.

      Delete
    3. I'm sure different professors put different emphasis on it which is fair enough. I don't think it's an esoteric insight, but it's easy to dismiss the problem as a mere technicality to be solved by clever people as opposed to a fundamental information problem.

      On Coase, I don't recall learning it as a counter-point to Pigou, but rather as just another approach to externalities.

      Delete
  2. That video made me laugh. You really can't fault libertarian's ability to propagate their cause.

    ReplyDelete
    Replies
    1. Did you really think it was that bad? I mean, yes Munger's politics come through very clearly but I thought he did a good job outlining the economics before getting into his political views and the transition from economics to applications was pretty clear.

      He did not say discussion of things like externalities are outside of the mainline economic tradition (the way Peter Boettke often does) or dismiss Pigou. I thought overall it was good, although it was very clear his politics were coming out in the applications.

      Delete
    2. Applying your politics is unavoidable. When I talk about externalities I talk about science funding and climate change. But that's because my politics are such that I think it's appropriate for the state to get involved.

      What was nice about this video was that he lays down the theory in a way that we can all agree with (I think).

      Delete
    3. I agree with Daniel here; to use an analogy, it was more like reading Econlog than Cafe Hayek :)

      Delete
    4. Perhaps my comment came across wrong. What I was saying was that libertarians are great at communicating their ideas in a light hearted, easy to understand and humorous way.

      This is in contrast with leftists, who are often frustrated or earnest, and hence easily mocked, or reactionaries, who simply come across as stupid and angry.

      Delete
  3. In my graduate school experience I'd say of the 10-20 professors that went over Pigou taxation, I can only recall 3 really noting the information problem. One was in passing (he said, "it's not a 'trivial' problem"), the other two in a reasonable amount of detail. Of the latter two, it was obvious that one had given it great deal of thought. For the other it was more so because of the course content (it was an environmental course, so policy is pretty much the sole focus).

    Ironically, the one who had given it the most thought - by far - was at the least reputable institution (and a Master's-only program at that). The balk of my professors have been at the other two schools, both very reputable, and both doctorate programs.

    This is only a small sample, but I really think there's a lot of truth to the statement. The information problem isn't really the focus of the "top" researchers - I get the impression it's not the path to tenure at the big research universities. And if you're not actually investigating something, I think it's natural to ignore it, or glaze over it at best.

    ReplyDelete
    Replies
    1. re: "The information problem isn't really the focus of the "top" researchers - I get the impression it's not the path to tenure at the big research universities."

      No, I agree with this. It's not the focus at all. I'd say that's because it's blindingly obvious (unlike the externality argument itself)! What would even lead someone to think the government would know this.

      I know we're really just stuck with dueling anecdotes here, but I firmly think it's a silly accusation.

      It reminds me of the people that make a huge deal of the socialist calculation debate. Yes, some economists like Lange thought with computers you might pull it off (using thoroughly neoclassical modeling principles, let's remember!). Yes the public was more enamored than they should have been with socialism. Maybe the SCD added value in combating both of those things. But the reason why few economists were talking about it and thinking about it as much as Mises and Hayek were was because most economists considered Marxism and socialism a joke. To the extent they supported socialist causes it was purely for ideological/egalitarian reasons and not because they didn't understand the economics behind market allocation.

      Delete
    2. I couldn't disagree more. I don't think it's blindingly obvious at all - the Soviet Union tried to run an economy using Lagrange multipliers. They were smart guys there, no? Most of the uninformed public still thinks that the main obstacle to Socialism is an incentive based problem - they have little idea about the information problems discussed by Mises & Hayek. And I don't think it's trivial to grasp.

      I really think you give a lot of these guys too much credit. I will grant that if you were to ask most of them whether they thought the government knows the correct tax, they would say NO. But my point is that most would say the inaccuracy is tiny, at least relative to the benefits. And the even bigger problem is that I don't think they've really grappled with the details - and that's what gets passed on to their students.

      In my opinion a solid course should spend weeks, if not a whole semester, discussing how you estimate a tax, how you estimate it's error, it's welfare effects, how reliable these methods are, etc. This is not trivial by any means, and it's extremely important, yet I think it's done very little. That really speaks to the importance given to the topic. You can say that it's blindingly obvious that the government would know the correct tax, but I don't think you can say the same for the question of "to what degree" and "what are the consequences".

      Delete
    3. Most of the uninformed public still thinks that the main obstacle to Socialism is an incentive based problem - they have little idea about the information problems discussed by Mises & Hayek.

      I think the same is true of most right-leaning/libertarian bloggers as well, though. I see so many use historical excerpts related to decimated agricultural yields and point to that as evidence that the Soviets and Maoists had no idea what they were doing; Anything having to do with information is much more generally described the same way it was in my sixth grade social studies textbook: "command economies can't respond to what people want like market economies can."

      I would also argue that the information problem in a Socialist approach is solved much more easily than the incentive problem, especially with modern technology and metrics. In other words, I think that a modern-day Soviet Union could absolutely have better control over its economy than the real one did... but it would still face incentive problems that would doom it just as readily. This to me means that incentives really are the core of the problem.

      Delete
    4. I would also argue that the information problem in a Socialist approach is solved much more easily than the incentive problem, especially with modern technology and metrics.

      It's comments like this that convince me that, counter Daniel, the SCD is not widely understood, or really appreciated.

      Let me put this way. If the Soviet Union had a population of robots - i.e. there was no incentive problem - modern computing power and technology would still not be even remotely powerful enough to enable successful central management of a large scale economy.

      I can probably be best stated in terms of modern Computational Complexity Theory. Like the "Traveling Salesman Problem" (a still infinitely simpler problem than what we're discussing), the type of coordination problem that would be faced would require more computing power than could possibly ever exist in the universe. It has to do with the size of the problem and how fast the necessary computing power grows in relation to that size. If the necessary computer power grows exponentially, modern computer power, even at the rate it's growing, would never be powerful enough to solve even moderately sized problems of certain complexity. It would take something ground-breaking - or put differently, it would take maybe the biggest discovery of the last century, to think otherwise. It all has to do with the P=NP problem (a very interesting problem - it's almost universally agreed that they're not equal).

      In any case, even calculating solutions to the largest (quite small) *basic* general equilibrium models can take months on powerful computers - and we're talking really simple models. You really have to opt for approximations, rather than solutions, and even so, we still would not be able to develop a model that would provide the outline for how to run an economy.

      Delete
    5. I agree that most people probably don't get the information/incentive subtlety that distinguishes the SCD. I have to focus my neurons to really put it intelligibly myself. That is just to say that a nuance that Hayek thought was important is lost on a lot of people. But we are not obligated to agree with Hayek on this. I don't think it's all that important.

      What I was saying is that it's not like people were in the dark about the problems with socialism before the SCD.

      And you don't need to make such a huge distinction of it that Hayek did. Incentives and information really aren't as distinct as I think a lot of people make them.

      Delete
    6. edarniw, I don't want to split hairs here, but you're saying that if the Soviet Union had a population of robots, you wouldn't be able to solve the information problem... except that you are positing a population of robots, which means you would be able to control them entirely, which means there wouldn't be an information problem to solve.

      I'm not saying that the information problem is easy to solve; I think you misunderstand my meaning here. However, modern technology and metrics enable us to collect data and make predictions that the Soviets could never have dreamed of, and the technology is getting better all the time. Yes, it would be absolutely impossible to predict an equilibrium with 100% certainty, but that's a standard of efficiency that is also impossible for an entirely price-based approach as well. There is no such thing as a perfectly efficient market in the real world, and I believe that with sufficient technology, it is at least conceivable that central planners could do a better job of planning and predicting with today's technology than they could fifty or a hundred years ago. Given enough information, it's conceivable that planners might even be able to approach allocation efficiencies somewhere near what a price-based market could accomplish.

      However, my point is that even with all of the information in the world, the incentive problem would doom such a project to failure. There would be no way short of absolute mind control that workers' incentives could be kept in line with central planners; it is my contention that this is where Socialism ultimately breaks down and becomes an unsolvable problem.

      Delete
    7. edarniw, I don't want to split hairs here, but you're saying that if the Soviet Union had a population of robots, you wouldn't be able to solve the information problem... except that you are positing a population of robots, which means you would be able to control them entirely, which means there wouldn't be an information problem to solve.

      Being able to control them has nothing to do with the information problem. The point of assuming robots is to get rid of the principal-agent problem, or the divergence between private objectives and the "common" objective.

      I can be able to perfectly control robots without getting any closer to solving the planning problem. It's what I tell the robots to do that constitutes the information problem.

      The problem of economic organization is that you want to organize all of the inputs to maximize some measure of welfare. There are an infinite number of ways to produce a given output. The question is how to do so in such a way that the inputs used to produce each output can not be rearranged to further increase welfare - economic efficiency. One way to get close to this is through a price system, but you only get prices when you have disparate entities bidding on items competitively.

      If there are not prices then you need to use something like Lagrange Multipliers - these now constitute the measure of importance of using an input in a given occupation. I can't begin to imagine how awful this would be.

      1. LM's (their ranking) are hyper-sensitive to the multitude of specifications of a model: technology, preferences, etc.

      2. We're talking a massively complex problem here. Imagine trying to decide how much labor to use to produce an input that will not become an output until five years from now. The number of stages between the use of the input and the output that it will be a part of makes it crazy complicated to figure out whether that input should be used there as opposed to another place.

      3. You're very limited to the size and level of complication of the model you use, even assuming approximation methods. Practically speaking, you'd be forced to limit the time horizon and substitutability of the processes and inputs in the model - the efficiency loss would be massive.

      These are just a few of the many critical problems involved when centrally managing a large pool of resources.


      And you're right, the price system isn't an optimal solution. But if you really grasp the complexity here, you'll concede that a planner wouldn't be able to even begin to approach the results of a price system. If we were to speak in terms of how big of a population each system could support, here is my ultra-speculative guess.

      A planning system with robots could not support anything more than in the thousands. A price system has already been shown to be able to support billions. Again, this is crazy speculative, but my point is to emphasize the magnitude in difference.

      Delete
    8. I agree that most people probably don't get the information/incentive subtlety that distinguishes the SCD. I have to focus my neurons to really put it intelligibly myself. That is just to say that a nuance that Hayek thought was important is lost on a lot of people. But we are not obligated to agree with Hayek on this. I don't think it's all that important...Incentives and information really aren't as distinct as I think a lot of people make them.

      I'm kind of shocked by these comments. You always give the impression that the Austrian contribution to the debate is obvious, understood, and highly appreciated.

      Mises' major contribution was to describe the role of the price system in solving the resource allocation problem. It has nothing to do with the incentive problem.

      And when you say that incentive-information differences are over-played I can only assume you're thinking more in terms of a principal-agent situation. In PA situations the incentive problem goes away if you have perfect information, or conversely, the info problem goes away if incentives are in line.

      But this is entirely distinct from the type of information problem that faces a central planner. Do you not see the difference? Is it not clear that the sort of problem one faces in optimally allocating resources is distinct from an incentive problem? Maybe the problem is more "computational" and I should be using that in place of information...

      Delete
  4. On the topic of externalities, I'm always a bit disappointed by the lack of discussion about its definition.

    I know there's been a good deal of research on it because it was discussed in some detail in the environmental course I mentioned in the post above.

    As in the video, the typical definition of externality entails a direct third party effect. Two of the obvious reasons it's defined that way is because it's mathematically stark and tractable.

    I wish more discussion would be given to the following chain of events - guy gets irritated from potato chip sounds and treats his friends awfully, or punches a whole in the door, or...you get the idea. And from those actions, other actions are affected, and other actions, and so on. By not considering the chain after the direct effect you're implicitly assuming that the rest is wash. In a complex dynamic system, I don't think you can always do that - at the very least, it needs to be given some hard thought.

    Of course, if you're throwing in every action of every person in every other person's utility function, you're not going to get very far. But that's kind of the point - you've really got to make a good case that the first-hand direct effect over-rides the following effects...

    ReplyDelete
    Replies
    1. I hadn't really thought of that before, and that makes a really good point. But it also seems to me that it is far easier to measure direct third-party effects, and that correcting for those might go a long way towards easing fourth-, fifth-, etc.- party effects.

      Delete
    2. But a lot times it doesn't go towards easing higher order effects. That was Bob Murphy's recent point on the tax interaction effect. For example, if you implement a tax on carbon, it will exacerbate the existing distortions from other taxes/subsidies. The empirical evidence so far indicates that on net that effect is negative. That is, more additional distortions are created than reduced, and it's not trivial either.

      That's the point of being more mindful of higher order effects - they are frequently in different directions, and sometimes systematically so.

      I should also point out that it's not just with the higher order effects that I'm concerned with. The typical definition of an externality is very narrow even ignoring that it tends to only account for direct third party effects. Even the equilibrium price that pops out of a model could in some sense be called an externality. I just think there needs to be a lot more attention paid to the alternative definitions that could be given, and importantly, what their consequences would be.

      Delete
  5. In the intermediate micro course I just completed, my professor specifically did mention the information problems associated with corrective taxes (although she did not call them Pigovian), and she did an entire lesson on Coase and property rights, complete with a mathematical breakdown. She also went into the information problems associated with Coase and the issues that arise when many people are facing the burden of the externality and must negotiate as a group. The video as a whole is good, but I think it suffers from a touch of the libertarian martyr complex :)

    ReplyDelete
    Replies
    1. Coase is standard, I will grant that. But that's quite different from the informational problem the government faces.

      And it's good that it was mentioned - but as I said, it should really get more attention, given how big of a role it plays in policy.

      Delete
  6. I used to blog on externalities a lot on here. It's a very important part of economics, and a lot of libertarians have a tendency to treat it (and Pigou) like voodoo.

    I think this voodoo charge is more a function of how non-libertarians believe libertarians think, than how libertarians actually think.

    Most libertarians know about the work of Rothbard and Block, who established quite clearly how private property rights prohibit externalities.

    Also, you'd be hard pressed to find a libertarian say what Coase said, which that the way to settle negative externality disputes is by granting permission to whichever party is offering something "more valuable to society"...choose your metric, than the other guy. So if I dump raw sewage onto your lawn, because of my raw sewage treatment operation, and you bring me to a Coasean court, then I could possibly win against you, if I can convince the judge that my operation is "more valuable to society" than what you are doing with your own home.

    THAT is voodoo to most libertarians, and even non-libertarians who have some semblance of moral rectitude, and lack of insanity.

    ReplyDelete
    Replies
    1. Rather than "prohibiting" externalities, I'd say that Austrians have reasons to believe that the market process tends to solve externalities, over time, better than other methods.

      Delete
  7. "Sure bargaining can solve some of these problems, but sometimes we simply don't think its right that person who bears a cost has to compensate the aggressor for damages done! In some circumstances we find that so distasteful we call it "extortion"!"

    It's a fun coincidence that you wrote this post the same day I was reading McCloskey's article on "The So Called Coase Theorem.
    http://www.deirdremccloskey.org/docs/pdf/Article_306.pdf

    You should take a look. She argues that one of the true lessons of the Coase article is that you can't (easily) make statements like the one above. If smoke from a local factory drifts into your house and makes your laundry dirty, it isn't obvious either party could accurately be described as an "aggressor" where property rights are unclear. You are both "responsible" for the pollution. The factory by emitting smoke and you by living close to a factory.

    ReplyDelete
    Replies
    1. Coase's point was specifically that this doesn't matter. You just assign property rights however you want and let the two parties trade.

      Delete

All anonymous comments will be deleted. Consistent pseudonyms are fine.