David R. Henderson  

Mankiw's Misleading Treatment of Public Goods

Apple's E-Books are Pro-Compet... How to Make Money: Go Where Ot...

In his treatment of public goods in Principles of Economics, 5th edition, Greg Mankiw gives the standard two characteristics of a public good: (1) the good is non-excludable, that is, a person can not be prevented from using it, and (2) the good is non-rival in consumption, that is, one person's use of the good does not reduce another person's ability to use it.

So far, so good. This is what, I suspect, almost all economists teach.

But then, to drive the point home, he shows these two dimensions in a matrix, with whether it's rivalrous in consumption on the horizontal and whether it's excludable on the vertical. In the "rivalrous in consumption but not excludable" cell, he puts, as an example, "congested nontoll roads." His idea is that if it's congested, it is rivalrous in consumption. That's correct. But he's saying the fact that it's a non-toll road makes it non-excludable. That's wrong. Excludability has to do with whether someone can be prevented from using it, not whether someone is prevented from using it. And by "can," I don't mean, and economists don't mean, the legal idea of "can" but the technological idea. It is technologically doable to prevent someone from using a road. So a road is excludable.

You might argue--and some of my students in the past have argued--that because a law says you can't exclude people, then that makes it non-excludable. But I point out that if that's the case, the government could make things non-excludable simply by not allowing people to exclude others. That's clearly not what economists had in mind in coming up with this characteristic. (It was Paul Samuelson in a 1954 article who stated this idea clearly.) What they had--and have--in mind is that there are no low-cost ways of excluding people.

In short, excludability is a technological characteristic, not a legal one.

My favorite example of a public good is a radio signal before scrambling was invented. It fit beautifully both characteristics of a public good: (1) your tuning into the radio signal did not diminish other people's doing so, and (2) it was prohibitively costly, before scrambling, to exclude people. One delicious aspect of this example is that it shows another thing that economists often want to show: namely, establishing that something is a pure public good does not establish that it can't be provided at a profit. Radio in the United States was, of course, provided at a profit for many decades. I first got this example back in the 1970s, from either Randy Holcombe or David Friedman.

Comments and Sharing

CATEGORIES: Public Goods

COMMENTS (26 to date)
Tony N writes:

I may be mistaken, but I think "rival" in the first paragraph should read "non-rival."

David R. Henderson writes:

@Tony N,
No, you're right. Change made. HT to Ted Levy also for private e-mail alerting me to this.

Philo writes:

“My favorite example of a public good is a radio signal before scrambling was invented.” Note that scrambling was always possible; it just hadn’t yet been widely figured out. Maybe a few people understood it, but they weren’t the ones running radio stations. So ‘technologically impossible’ doesn’t mean ‘technically (physically) impossible’: it means something like ‘impossible using the techniques that the agents involved already know about, or could easily find out about’. In short, it’s quite an imprecise concept.

Austin writes:

A road is excludable, a non-toll road is not. You're changing the good.

Russ writes:

Clean, breathable air was the example that sticks in my mind.

Thomas Snyder writes:

A non-toll road can be excludable if it is in a gated subdivision.

mike davis writes:

True, it may be profitable to supply a pure public good like radio or lighthouses, but that doesn't mean we will get the optimal supply. Despite what Garrison Keeler might tell you, the glory days of radio were not nearly as interesting as today's world of XM, Pandora and Econtalk.

I agree with pretty much everything in this post. However, if the definition of excludability is "that there are no low-cost ways of excluding people", I would not limit the definition to being only a technological one. The cost of available technologies depends on the value individuals within society ascribe to them. So, for example, if it is technologically possible to exclude someone, but people choose not to because it is too expensive, that means excludability is not only a technological characteristic, but also a characteristic that depends on preferences. Subjectivists like Mises would probably go even further and claim that it is entirely up to our preferences whether or not something will be a low or high cost alternative.

Another thing to consider is that the legal environment and cost can't be disentangled. Costs of performing various exclusion activities depend on our expectations of what other people consider as an acceptable exclusion technique.

Finch writes:

I've wondered about the analogy between radio spectrum and roads before. Seems to me that a bag of nails on I-95 is not too different from a noise broadcaster on a piece of spectrum. Bad guys can easily exclude you from either. The radio troublemaker would even be easier to catch.

I don't think this is a particularly good justification for the draconian regulatory regime in radio.

Daniel Kuehn writes:

re: "One delicious aspect of this example is that it shows another thing that economists often want to show: namely, establishing that something is a pure public good does not establish that it can't be provided at a profit."

This is a tricky point to communicate well too, though, because some people confuse providing something in a market setting with providing it optimally in a market setting. Roads are a great example of this. Lots of people invoke private road building as a retort to the case for public roads, but that's only half the argument. Most advocates of public roads don't think zero roads would be built without the government. The question isn't just "are private roads profitable enough to be build" - it's also "will enough roads be built privately?"

James writes:

Daniel Kuehn,

You write: The question isn't just "are private roads profitable enough to be build" - it's also "will enough roads be built privately?"

Excellent point but still something is left out. The public sector or the private sector can reasonably be expected to build some number of roads and in both cases the number of roads will differ from whatever is optimal after all costs and benefits are taken into account. The question is which option is least bad.

Most people seem to believe that the government will come closer to the optimal amount of road building but I've never seen any evidence for this. Eisenhower built the US interstate highways to move weapons in anticipation of a US vs USSR conflict that never happened. Many presidents have embraced infrastructure projects for the purpose of giving the unemployed something to do. Congress has used federal highway money to encourage lowered speed limits and drinking ages. State and local governments do a lot of road building near fiscal year end to prevent reductions in future year budgets. And it seems that the potential costs of climate change from burning fossil fuels are being revised upward all the time. Given the available facts, would you rather that road building was left the the private sector or would you rather it be a government program?

awp writes:



market urbanism brought this point up a few years ago, which led to an interesting discussion of the difference between marshmallow and lollipop highways.

PrometheeFeu writes:

With regard to non-excludability, I think it depends on the context. If the government says you cannot exclude anyone, then the good will sometimes behave as if it was non-excludable. So for the purposes of the analysis, dismissing the government's intervention is counter-productive. You'll make bad predictions.

If you looks at the market for music for instance and ignored copyright law, your analysis would be wrong. (of course, the same would be the case if you ignored piracy)

However, if you are trying to calculate optimal policies, then of course, policies make for bad inputs to your model.

John writes:

@Danniel Kuehn

But following Tulluck (57 I think) we also need to ask "Won't too many roads be build publicly?" Moreover, we also need to ask if the economic cost of the roads is not too high for what we get from the marginal road (or marginal quality of the roads over what might be provided privately).

I'm not really disputing the point you raise about why societies seem to turn road building over to government at some point. I do question whether the public goods explanation is accurate.

Daniel Kuehn writes:

John, James -

Right, it's the same "will it be optimal"? Question I posed. You can't just point out a government failure or a market failure - you've got to think about the competing options.

James asks what I think makes the most sense. I agree with John (for the reasons David states here) that it's not a public goods argument that justifies public roads. It's an externality and network effects argument as far as I'm concerned. We seem to agree that private roads provision will be suboptimal, so we should agree (I think?) that modest public provision will move us closer to optimality (unless you think the government is horrendously inefficient at building roads - I would say that's unrealistic). So the next question is - does it overshoot? Does it produce too many roads and overshoot the optimal point? Are there areas where there are a lot more public roads than there are actually used is a good indicator of that. Nowhere where I live. The most unused roads are going to be long connector highways on the interstate but presumably they're more than justified as connectors of major nodes. Is there any evidence that we've overshot or that if we have overshot we've overshot so badly that we are farther away from optimality than having no public roads?

I don't know of any evidence of this but maybe someone knows better.

New public roads are added to state highway systems presumably using different rules in different states - I know in Virginia the state highway system's rules are pretty clearly targeted toward ensuring that new state highways generate network effects. Those are exactly the sorts of rules you would want in place if you want to generate a near-optimal number of roads.

Any reasons to think we've screwed ourselves on the whole public roads thing?

I know that Mankiw is aware of the argument, because I have pointed it out to him, and he's acknowledged it.

For about eight months, drivers in 275 Seattle-area households agreed to pay for something the rest of us get for free: The right to drive on the region's freeways and streets.
They were guinea pigs in a pioneering study that explored how motorists' behavior might change if they had to pay tolls — not just on a few bridges or highways, but on almost every road with a yellow center line. Researchers established virtual tolls ranging from a nickel to 50 cents a mile. They gave participants pre-paid accounts of between $600 and $3,000, and told them they could keep whatever the tolls didn't eat up.

I've also pointed out to David Friedman, when he brought up the radio broadcast example in a debate, that what entrepreneurs did was to take a pure public good and make it the raw material for a private good; an audience of listeners, the right to broadcast advertisements to COULD be sold.

An amusing example of privatized roads comes from Chile. Ricardo Lagos, a member of the Chilean Socialist party, when he was Minister of Public works in the 1990s there, realized that the Pan American Highway that is the main artery the entire length of Chile, needed to be repaired. He leased it to entrepreneurs who did just that, and recovered their investment through tolls.

'Necessity, mother of....' Even for descendants of Salvador Allende.

Btw, Mankiw's new paper is positively Friedmaniacal. He concludes that the evidence probably shows that CEOs earn their high incomes by providing good value, that the 1% more than pay their fair share of government 'infrastructure' costs, and the John Rawls 'veil of ignorance' argument for 'redistribution insurance' is pretty silly:

Yet take this logic a bit further. In this original position, people would be concerned about more than being born rich and poor. They would also be concerned about health outcomes.
Consider kidneys, for example. Most people walk around with two healthy kidneys, one of which they do not need. A few people get kidney disease that leaves them without a functioning kidney, a condition that often cuts life short. A person in the original position would surely sign an insurance contract that guarantees him at least one working kidney. That is, he would be willing to risk being a kidney donor if he is lucky, in exchange for the assurance of being a transplant recipient if he is unlucky. Thus, the same logic of social insurance that justifies income redistribution similarly justifies government-mandated kidney donation.

Paul Krugman should be apoplectic.

James writes:

Daniel Kuehn,

You are treating utilization as evidence that a road is not excessive. To see why this is wrong, consider the hypothetical world where road building costs so much that the optimal amount of road construction is zero. If the government builds a road anyway, say to move weapons or create jobs, the public is worse off but the road would probably be well utilized so according to your choice of indicator, there would be no indication of excessive road building. Any analysis based on utilization is therefore going to be biased in favor of excessive road building.

Now if you have an estimate for the optimal amount of road building, I'd love to know about it. Absent such an estimate, we can only rely on what we know about how public officials make road building decisions. My prior comment listed a number of forces that would lead to excess production so that's why I'm inclined to believe government production is going to be excesssive. Finally, the net externality of road building may well be negative if increased roads lead to increased fossil fuel consumption which leads to costly changes in climate. I have no idea why environmentalists haven't picked up on this yet.

I'm curious about the rules you mention in VA. What is the penalty that government officials face if they build a road that does not generate meaningful network effects? Is it severe enough to discourage excessive road building? For context, what is the penalty for going 10 mph over the limit in VA? Do people still speed?

Daniel Kuehn writes:

James -
You're ignoring the part where I talked about building roads being horrendously inefficient to build (or am I confusing how that's different from "costs so much that the optimal amount of road construction is zero". I don't consider that to be a realistic scenario. If you do that's fine, but I didn't ignore it. Do you honestly think that's true? Why?

Public transportation costs are essentially funded by user fees, so utilization seems to me to be a decent indicator of the balance of marginal benefit and marginal cost of the road infrastructure. If marginal cost far exceeded marginal benefit you'd expect people to be opting out of road use. I don't think you see a whole lot of that.

In the absence of a market (that's the situation with externalities unfortunately) we've got to rely on things like that. Do you have a better indicator?

re: "I'm curious about the rules you mention in VA. What is the penalty that government officials face if they build a road that does not generate meaningful network effects?"

No, private roads are only accepted if they meet requirements about public use, connecting a certain number of addresses, connection to an existing state road, etc.

I think I had over a $100 ticket once - it's something like $5 for every mph over plus a flat court fee. I can't imagine how this relates to road building. You've lost me now.

James writes:


My high cost hypothetical proves the existence of a case where utilization is a terrible indicator of excess production. I don't think the social cost of roads is actually so high. I just chose that to show why utilization is an absolutely unreliable indicator of excess.

Here's a more real world explanation of the same point. Road utilization indicates that the incremental private benefits to drivers exceed the incremental private costs that the drivers bear for getting on the road. But what's relevant is the total cost to society and the total benefit and no one chooses to drive based on total social cost and total social benefit. If VA builds a road that costs more than its value to society, people will still drive on it if their personal cost of getting on that road is less than their personal benefit of getting on it. Ergo, utilization is useless for determining if a road generates a net social benefit.

On the other hand we do know plenty about how policy makers allocate funds to road building. They are not even trying for anything like social optimality. Policy makers have a host of reasons to favor excess production and there is little or no countervailing incentive to avoid excess production.

Contrary to you claim, public roads are not funded by user fees in any meaningful way. They are funded out of general revenues just about everywhere in the US.

The price of a speeding ticket relates to road building in this way: You cite rules in Virginia that road construction should target network effects. In order for those (or any) rules to be relevant, they would have to affect behavior in a meaningful way. Since people still speed, we can safely say that the the fines on speeding are too low to meaningfully affect behavior. So unless politicians face a significantly greater fine than $100 for building excessive roads, the rules you cite are unlikely to actually prevent excessive road building. What penalty do politicians actually face in VA for building roads that don't generate large network effects? If you found out that there is no penalty at all for excessive road building, would you still expect VA's rules to influence the amount of road building?

Daniel Kuehn writes:

re: "My high cost hypothetical proves the existence of a case where utilization is a terrible indicator of excess production. I don't think the social cost of roads is actually so high. I just chose that to show why utilization is an absolutely unreliable indicator of excess."

Being "absolutely unreliable" and being "unreliable" are two very different things. Let's not have the latter piggy-back on the former (particularly when I made no claims at all for absolute reliability and even noted that it's not a perfect metric.

Do you have a better metric?

That's part of how I make the assessment in this imperfect world. How do you come to your conclusion?

re: "Since people still speed, we can safely say that the the fines on speeding are too low to meaningfully affect behavior."

What are you talking about? 364 days of the year I keep it within ten over the limit. Most people do. How in the world do you come to the conclusion that speeding tickets don't impact behavior? You should drive with my wife in the passenger seat - then you'll now how much they affect behavior. This doesn't seem like a very serious treatment of the issue.

Politicians look pretty closely at where the unmet demand for transportation is. Sure there are boondoggles and airports for guys like Murtha strewn across the country, but that's not where the bulk of the spending goes. The principle disciplining device is the voters. People vote on the basis of things like roads when it comes to state-level elections. And they are building roads in Northern Virginia, Norfolk, Virginia Beach, etc. at much higher rates than elsewhere in the state. Care to guess why?

You are taking good solid public choice ideas but you are assuming they drive everything rather than understanding how they operate on the margin. You can't just assume that - you have to justify that.

I think public choice offers a great framework for thinking about the problems with public transportation on the margin, but I can't just substitute that for an argument for why spending has to be more than optimal. I have to have a good reason for thinking it's more than optimal and it seems to me you don't.

James writes:


Neither you nor I have any metric to indicate what is the optimal amount of roads. Utilization doesn't measure that variable in question and your statement that roads are funded by user fees is a falsehood. Please look at the steps you have taken to defend your position. This should tell you something about your position.

You write "I have to have a good reason for thinking [public road building is] more than optimal and it seems to me you don't." I want to be 100% clear since you keep missing this: I reach the conclusion that governments will engage in excess road building by looking at the available facts concerning how politicians have made road building decisions historically. I have said as much twice already.

Of course speeding tickets have some effect on behavior, but violations occur, right? Now what is the penalty that politicians face for building roads that do not improve social welfare? If it's zero (far less than a speeding ticket, yes?) why would anyone expect the rules of VA to prevent excessive road building?

Public Choice, is not a substitute for an argument for why spending is more than optimal. It is such an argument. In short, politicians obtain personal benefits and the public bears the cost from excess road building. Politicians are people. People do things that get them benefits if the personal costs are less than the benefits. Why do you think the rules in VA will be effective at preventing excessive road building?

Daniel Kuehn writes:

OK, for the last time the public choice argument is no more a demonstration that roads will be overbuilt - it is a reason why they could be overbuilt. It does not work like externality arguments where there is a benchmark so that you can say definitively that you are going to get sub-optimal production.

Public choice offers a reason why it could be overbuilt, but not for why it will. You need more pieces to the argument. The benefits of things like this are about as diffuse as the costs and there's nothing like the certainty that you're assuming.

If you can't come up with a different argument don't accuse me missing the one you've made or abuse public choice theory to get the answer you want.

Daniel Kuehn writes:

I've mentioned penalties.

But let's say there are zero penalties. What do you expect politicians to do exactly? How does building a road benefit them so much that you're so confident they'll overbuild it? Voters don't like spending and they don't like deficits. The benefits of road building are not particularly concentrated - they are widely diffused. Exactly how does it follow that there is an incentive to actually overbuild even if there were no costs?

Public choice isn't just some kind of grab bag you can reach for when the government does something you don't personally like.

James writes:


You ask: "But let's say there are zero penalties. What do you expect politicians to do exactly?" I expect politicians to make road building decisions based on a collection of multiple motivations besides social optimality. Most, if not all, of those motivations will call for building more roads rather than less. I expect politicians to do this because politicians have been open about making road building decisions this way in the past.

After reading the preceding paragraph, do you still not know what I expect politicians to do and why? If not, I'm not sure I can help.

I agree that public choice is no grab bag. It is a framework for anticipating what people in government are likely to do. That's how I'm using it. Accusing me of using public choice as a grab bag doesn't even address, let alone rebut my argument.

Comments for this entry have been closed
Return to top