Wednesday 20 June 2012

Externalities: A Primer

It is far too easy to come away from a lazy version of Econ 101 thinking that, because externalities are pervasive, the range of meliorative government policy is very broad indeed. It's hard to think of any part of life that doesn't involve external costs of one sort or another.

When you get past Econ 101, you start getting all the caveats. The range of externalities that actually generate inefficiencies is pretty small relative to the universe of externalities. And, even those that can potentially be fixed by policy aren't necessarily best handled by policy. Private solutions to some kinds of externality problems do emerge; layering policy on top of private solutions can make things worse than leaving things alone.

Let's start by walking through a narrative version of Buchanan and Stubblebine's seminal work. And, we'll start by separating out two very different kinds of ways that other peoples' behaviour can affect you.

Suppose that you and I both attend a house auction. I show up wearing a t-shirt that is so offensive that you feel physically ill.* In fact, you would have been willing to pay me $100 to have worn a different t-shirt this morning. I would have been willing to have changed shirt for any payment greater than $20. Despite your nausea, you are able to bid on the house. But, because I'm bidding against you, you wind up paying $10,000 more than you otherwise would have.

My bidding against you at auction imposes a cost on you, but one that does not change the efficient outcome. The house goes to the person who values it most: you. And while you have to pay more for it, the seller receives more for it: my action has caused a transfer from you to the seller. When we model things more formally, these effects run through your budget constraint rather than your utility function. You have less money with which to do things, but we have no reason to expect that anybody is choosing the wrong mix of goods given their budgets. We call these budget-affecting externalities "pecuniary externalities". When economists say, "Oh, but that's only pecuniary", that means that it's something that we don't really need to worry about if what we care about is efficiency. We only care about externalities because they induce resource misallocation; pecuniary externalities don't do that. There might be other justice-based reasons for worrying about pecuniary effects, but to the extent that we do, we generally conclude that those kinds of concerns are really best handled by just giving money to poor people.

But the nausea that I imposed on you with my careful choice of t-shirt is a real external cost. It's easy to imagine that you'd be willing to pay some amount to change the state of the world - that means it's potentially Pareto-relevant. A Pareto-efficient outcome is one where it's impossible to make anybody better off without making somebody else worse off at the same time; a Pareto-improving move is on that makes at least one person better off while making nobody worse off. A Pareto-relevant externality is one where a Pareto-improving move is available: the total willingness to pay to change the outcome, aggregated across all the people affected, is enough to generate a change in the outcome. You suffered costs of $100; I would have been willing to accept $20 to avoid imposing those costs. Eighty dollars worth of value was forgone by my having worn the wrong shirt. So the externality was Pareto-relevant: had I chosen a different shirt, the gains to you would have outweighed the losses to me. We call externalities of this sort technological: they affect others directly through their utility functions rather than indirectly through the budget constraint.

We expect technological externalities may induce inefficient outcomes. The real costs that I impose on you through your utility function are not factored into my decisions and so I can wind up choosing the wrong consumption bundle. But, again, not necessarily. If it would have cost me $100 to change shirt, and the burden on you was only $20, the efficient result is that I wear the t-shirt. The externality was only potentially Pareto-relevant, not actually Pareto-relevant.

Eli Dourado's Mercatus Working Paper on cybersecurity walks nicely through the difference between inframarginal externalities and marginal externalities, along with a host of policy and private solutions to externality problems. David Friedman's treatment is excellent; he also cautions in Law's Order of problems that can arise when we couple Coasean solutions to externality problems with tax-based approaches [see Chapter 4 in particular].

What about externalities that affect you through the tax system? Edgar Browning calls these "Fiscal externalities". They would count as pecuniary by the Buchanan and Stubblebine definition, but they also are effects that operate outside the market process. David Roberts worries about these when weighing the case for taxes on sugary drinks. In the presence of a public health system, any decision you make that affects your health also has effects on others through the tax system. To what extent is it efficient to worry about them?

First, imagine the limiting case in which nobody's behaviour changes as consequence of health care costs being borne by others: everyone behaves as though they had private insurance that charged actuarially fair premiums. In that limit case, every health cost borne through the tax system is only a transfer. It's a transfer from richer and healthier people to poorer and less healthy people, and it's a transfer from risk-averse to risk-preferring people, but it has no efficiency consequence. There is no simple efficiency case for worrying about it, though you could build complicated cases around the cost of raising money via taxes versus the costs of abating health care costs via other mechanisms.

Now, let's take the more realistic case where individual behaviour is at least somewhat responsive to that somebody else is paying for your hospital care and that your premiums do not vary with your costs. There's reasonable evidence for this: Jon Klick and Thomas Stratmann show that when state governments mandated that private insurance plans cover substance abuse treatment, people increased their alcohol consumption by an amount equivalent to about "48 extra beers per person per year".

What are the policy implications? Let's recall first that the cost imposed by the behavioural change is likely to be small if consumption changes are small relative to the total amount that would otherwise have been consumed: average per capita beer consumption in the US is 78 litres according to Wikipedia. Imagine that we had no public health insurance but instead we gave a lump sum of cash to everybody and required that they use it to purchase health insurance; insurers could charge what they liked. How much would we really expect anybody to cut back on their consumption of sugary drinks? People bear very direct personal costs of being in poor health; the monetized value of that will not be trivial even relative to expensive insurance premiums. In other words, the effect may mostly be inframarginal. I refrain from engaging in extreme mountain biking even though ACC would cover all of my health costs: I really don't like pain and injury. I would pay multiples of the cost to the health system of a broken arm to avoid breaking my arm.

So any policy seeking to internalise the external costs of individual health-affecting decisions ought to focus on the incremental costs caused by the extrernalisation of those costs through the public health system, not the aggregate burden of related diseases. Browning, in the piece above-linked, shows that the set of taxes and subsidies that would be necessary to nudge everyone back to what they would have been doing if the health system didn't distort behaviour winds up replicating the insurance premiums individuals would have been paying under private insurance but at very high administrative cost. Private insurers can set premiums based on individually specific characteristics; governments are more typically constrained to using linear tax schedules. But for things like sugary drinks, health costs are likely exponential in consumption, not linear; for alcohol, it's a J-curve with health benefits from moderate consumption and costs from excess consumption. Your insurer might decide only to increase your premiums if you're a heavy drinker; governments are restricted to linear taxes on alcohol that overcharge moderate drinkers and undercharge heavy drinkers.

Setting excise taxes to try to internalise the largely pecuniary effects of consumption decisions can induce technological externalities by causing some would-be moderate consumers to consume too little.

A few bottom lines study notes for the exam, as this will also now be part of the readings for my Economics & Current Policy Issues class:
  • There is no market failure case for responding to pecuniary externalities. 
  • While there can be a market failure case for policies addressing technological externalities, we have to be awfully careful because private markets often have already internalised them. A decent rule of thumb: if people are linked through a contractual or quasi-contractual nexus, intervening is likely a bad idea. We don't need policies to address the costs screaming babies impose on other airline passengers; the airlines are residual claimant.
  • Most of the "cost to the taxpayer through the public health system" aspect of individual consumption is pecuniary rather than technological. Any policy in this area has to be very careful to address only the parts of consumption behaviour that are directly caused by the potential for public defraying of health care costs; the risks of inducing inefficiency by setting policy to address pecuniary effects are real.
  • Note too that the actual burden on the health system may be pretty complicated. Smokers cost the health system more in every year that they are alive, but potentially save the health system money in the long run by dying early - and they definitely save the government money in total when we consider that smokers die before drawing too much in superannuation. It's even pretty plausible that healthy people wind up costing the health system more in total because they impose costs for a long period of time. What do we do if we find out that it's the gym rats who wind up costing everybody more because they spend ten years in a publicly funded dementia ward instead of dying of a heart attack at 67?
This is part 2 of a two-part series; in part 1, I took a more rights-based approach. And see here for prior posts on externalities.

* Only weirdos would be offended by these fine shirts, but they count in the social welfare function too. Anybody stumped for Christmas presents for Crampton can send me one of these shirts; I'm likely a medium.


  1. Excellent post. Note that bottom line #2 above implies that there is a far stronger case for banning smoking on a public park bench than for banning smoking in privately operated bars and restaurants. The establishment internalizes the externality through its p&l.

    1. True. The actual health risks to others from smoking out in the open air are pretty trivial though; the case would likely have to be based on people's distaste for having smokers around. And then we get into what sorts of prejudices ought be given policy effect.

  2. Maybe it is only little me, no idea of the externality costs of my life and here in Bangkok the external wife costs are heavy, and that what I say dudes, if Eric can get down these external wife costs well I say he can Canadian go for it in Christchurch, we need a good boss for University there, I said to Brownlee you sack everybody in Christchurch ow dude, and Brownlee said to me you got a good idea dude, but you shut up you,

  3. Eric,

    I'm not convinced that "There is no market failure case for responding to pecuniary externalities."

    Say a set of firms collude and raise their prices above the competitive level, then the effects are totally mediated via price (it seems to affect budget constraints rather than utility functions) but the action of the cartel causes aggregate economic harm. There is no effect on production or consumption, in this case, that is not caused by the price rise. Surely there is at least may be a case for responding to this pecuniary externality? Courts awarded compensation in common law for such actions.

    More generally, although it is true that pecuniary externalities do not affect the optimality of competitive markets, if markets are distorted (eg by taxes or monopoly), then shifts in supply and demand curves have efficiency effects.

    1. Mark: I'm pretty sure that standard theory says cartels are bad not because of the pecuniary externality but rather because of the Harberger Triangle. Then you get into all the Tullock stuff about competing to get the position.

    2. “Mark: I'm pretty sure that standard theory says cartels are bad not because of the pecuniary externality but rather because of the Harberger Triangle. ”
      Yes, that was my point. Something that operates through budget constraints and sees to be a pecuniary externality on your definition can have efficiency effects.
      For example, take a market distorted by a tax, which drives a wedge between the value of the product to the buyer and its value to the seller. Shifts in demand and supply curves in that market have efficiency effects (represented by the change in revenue from the tax in that market). Likewise in a monopoly distorted market. And imposing a tax or creating a monopoly has pecuniary effects only, but creates a deadweight loss. (Incidentally, I thought Brownings analysis of fiscal externalities raised the issue of whether there was some reason for subsidising medical markets that we forget to account for in the standard analysis).
      Here is another example. Take Friedman’s analysis of immigration (pp.423-25 of my second edition – figure 14.5). when we add an immigrant, shifting the supply curve, there are pecuniary externalities on existing suppliers and demanders in the market. But, as Friedman shows, the gain to demanders is bigger than the loss to existing suppliers – there is a Harberger triangle gain to the existing residents. The pecuniary externalities do not offset. That is the argument for immigration – that it leads to an efficiency gain for natives as a whole. (In practice, there may be other pecuniary effects through the social welfare system – and also technical externalities if immigrants directly enter our utility functions through cultural effects).
      Pecuniary externalities are offsetting in competitive markets with a fixed population, but not otherwise.

    3. I would have modelled the cartel arrangement as running through the production function rather than the cost function.

      Greenwald & Stiglitz do find more generally that pecuniary effects can matter in second-best worlds.

  4. Eric,

    I think this was an excellent post. Translating this into policy, however, I see two problems:

    1. the distinction between pecuniary and technical externalities is a difficult concept to explain even to people with some economics background; it is likely to go way over the heads of many policy-makers, let alone the public.

    2. more importantly, I think allocative efficiency is not the motivating factor behind healthist policy interventions, even when “market failure” is cited as an economic rationale. What most people – and I think many policymakers – are aiming at remedying is the pecuniary or fiscal externality. The common argument you will hear from the man in the street in favour of healthist interventions is “I don’t care what you do, as long as I don’t have to pay for it.” You never hear random people say “lack of a soft drink tax leads to inefficient societal resource allocation.” What motivates people to support soft drink taxes is that they have to pay for the health costs (either through higher premia or higher taxes) of people who eat drink too much of the stuff. So arguing against healthist policies on allocative efficiency grounds isn’t going to work as a practical matter.

    1. On (1), that's why I think an awful lot of places need to be more careful in how they teach externalities. We should be focusing far more on the misallocations rather than the "somebody else pays" aspect.

      On (2), agreed. And that's why I get rather ... exercised about social cost studies that purport to present costs that users bear as being instead borne by others. But I do think too that pointing to externality arguments gives a sciency veneer to it all. It's exactly what Sir Geoffrey Palmer did when he explicitly made a Pigovean case for increased alcohol excise/regulation on the basis of the difference between social cost and excise revenues.

  5. A serious problem with the use of externalities in policy arguments is that things often have both positive and negative externalities, as you mention, and that their size may be uncertain. So if you want to ban or tax something, you make generous estimates of the negative externalities, conservative ones of the positive--and perhaps miss some of the latter, since you aren't really looking for them. You then conclude that you have an objective argument for the policy you want. Similarly in the opposite direction if you want to subsidize or require something.

    In my experience, this is a serious real world issue. I've discussed it at some length in the global warming context on my blog, for instance at:

    For a much earlier discussion in the context of population issues, see:

    1. There's also the wonderful Tullock piece on externalities in government. Public Choice, 1998 I believe.

  6. And Kiwi Dave is correct about the difficulty in explaining the concept. I’ve just been puzzling over the some law and economics textbook treatment of the reluctance of the courts to award damages for the pure economic loss resulting from a tort. Many law and economics texts claim that is because they are pecuniary externalities and should not be awarded on efficiency grounds. But it isn't clear to me that is a relevant explanation.

    For example, in Rickards v Sun Oil Co., the defendant’s negligence put the only bridge between an island and the mainland out of commission. Merchants on the island who saw their business dry up as a result of the collision sued the defendant but lost, the court holding they were not allowed to recover ‘pure economic loss’ from customers who would have bought their goods if the bridge had not been destroyed.
    Posner (and other texts) writes that this was “the right result. Although they did lose money, most and perhaps all of their loss was a gain to mainland merchants who picked up their business when customers could no longer reach the island. Since the defendant could not seek restitution from the mainland merchants of the gains he had conferred upon them, it would have been punitive to make him pay the losses of the island merchants. The net social cost was damage to the bridge.” Dnes even has a little diagrammatic analysis where he shifts demand curves, wrongly shades in the areas between them as change in consumer surplus and the result is somewhat confused (but rationalises the court decision).

    That explanation seems totally wrong to me. When the destruction of the bridge reduces supply to mainland consumers and drives up the price on the mainland, the gain to mainland firms comes from their customers, not the island resort. Surely the pecuniary external benefit to the mainland merchant is (more than) offset by a pecuniary external cost to the mainland consumers. Take the simple case where there are no island consumers and the producers produce the same good. There is a net loss to mainland producers and consumer combined, so nothing (less than nothing) is left to balance the loss to merchants on the island. (Actually, this is just Friedman's the analysis of an immigrant coming in and adding to the supply curve in reverse).

    If the Posner analysis were correct, I could fence in a business, or through a picket line up, and say there was no social damage.

    What if the bridge were privately owned and charged a toll for crossing? Surely the owner could sue for lost toll revenue. And the island merchants sue for lost surplus.

    Also the destruction of an essential input into the island merchants providing for the customers seems to me to be a real externality, not a pecuniary one. One thing I like about law and economics is how it forces you to apply theoretical concepts to practical real world problems (‘concretize’ them as Friedman says), which usually demonstrates how useful are the concepts. Except perhaps here it reveals I'm not sure what counts as a pecuniary externality. Also, what if the firm was vertically integrated, what would be a pecuniary externality if they were dealing thjrough contract becomes something that affects the firm's production function.

    1. I pointed out the error in Posner's analysis of Ricards in my essay on him in the New Palgrave. According to the account in that essay--it's been a long time so I must rely on my past writing instead of my memory--the error was pointed out still earlier by Mario Rizzo and accepted by Posner in his fourth edition. I haven't checked the current edition to see how it treats the case.

  7. (Continued from previous - I was over the word limit)

    In Weller & Co. v. Foot-and-Mouth Disease Research Institute an infection of foot-and-mouth disease escaped from the research institute as the result of their negligence. Cattle in the surrounding areas were affected and the Minister of Agriculture, acting under statutory powers, closed the local cattle markets in which the claimants carried on business as cattle auctioneers. Since the claimants owned no cattle or other property at risk it was held that no duty of care was owed to them in respect of their loss of business profits.

    Veljanovski writes
    “The loss was a pecuniary externality. The negligent act of the defendant did not directly affect Weller’s production activities, but only led to a change in the market for their services – a fall in demand, which was similar to falls caused by other fluctuations in the market resulting from non-negligent factors. The social cost was the technological externality of harm suffered by the farmers, whose loss in physical production would be met by the undisputed liability of the defendant towards them. The compensation paid to the farmers preserved their incentive to raise cattle in the future, and the Institute’s liability to them provides incentives to similar research institutes to take care. Since the farmers’ real rate of return was unaffected the auctioneers suffered no permanent or long-run diminution in the demand for their services or their productive capacity.”
    If it was a pecuniary externality coming from a shift in the demand for their services (to zero), then presumably the loss to the auctioneers is balanced by a gain to their customers, the farmers (they no longer have to pay the auction fees). But wouldn’t that mean the defendant no longer has to pay these losses – as that reduces the damage suffered by the farmers and the damages payable to them. That is, if it was a pecuniary externality, it may still be desirable to charge the defendant that loss. Sometimes a pecuniary externality is offset and irrelevant, but it may be relevant for determining who lost as a result of the defendant’s negligence.

    The textbooks seem to be rationalizing these decisions by saying it would be inefficient to charge a pecuniary externality, but is that true? I would have thought the reason for limiting liability is some notion that the victim is the least cost avoider, and not paying damages gives the victim an incentive to minimise these losses, both exp post and ex ante (reduce reliance investment), or they perhaps can insure – although the courts seem to think a problem with pure economic loss is its unforeseeability, which would make it difficult to insure). Or maybe the courts are worried about difficulties in estimating the lost surplus.

    Further, is this really an example of a pecuniary externality? The defendant wiped out an essential input into the plaintiff’s business.

    And if the auctioneers and farmers were vertically integrated, could we then say it directly enters the firm’s production function and so is no longer a pecuniary externality? If we are to limit liability for pure economic losses, then whether it is a pecuniary or technical externality doesn’t seem that relevant.