Adding a road can make everyone's journey slower. In the mathematical network studied by Dietrich Braess, drivers choose whichever route offers them the shortest journey under the traffic conditions they encounter. A new connection changes those choices. Traffic redistributes, congestion develops differently, and the resulting equilibrium can leave each driver worse off. The extra road has increased the options available while worsening their combined result. This happens under particular network and demand conditions; it is no general argument against road building. It shows how the behaviour of a system can depend on relationships that individual decisions fail to capture. 1
A driver considers the delay on a possible route. Entering that route can also delay other drivers. Similar gaps appear wherever an economic decision changes conditions outside the agreement authorising it. Imagine a workshop buying a chemical, using it in production and releasing a residue that damages nearby gardens. The supplier and workshop may be satisfied. The gardeners have agreed to nothing, yet they inherit some of the expense through lost crops or the work needed to restore their land.
We call this an externality because the effect falls outside the relevant transaction. The word identifies a boundary in the account; it says little about the importance or physical distance of the consequence. Effects can also be beneficial. Useful work may improve conditions for people who have made no payment for it. The issue is how we identify these changes and make responsibility intelligible when the person deciding and the person affected occupy different parts of a complicated system. 2
Arthur Pigou approached the problem through divergences between private and social net products. An activity can yield a private return while imposing uncompensated costs on others, leading private incentives away from the socially desirable outcome. He considered taxes and subsidies among the means of narrowing that divergence. A corrective charge can make a previously cheap damaging method less attractive and give a firm a reason to change its process. Its corrective purpose is to alter the decision at the margin; revenue alone tells us little about that effect. 3
That approach requires information. An authority needs to assess the relevant damage and the response a charge is likely to induce. It can set a weak incentive that leaves deterioration profitable, or impose an expensive rule poorly connected to the problem. Collection and enforcement also cost resources. These difficulties deserve examination alongside the damage we hope to prevent. They do not establish that public intervention is necessarily less effective than every available private agreement.
Ronald Coase insisted on comparing such arrangements. His analysis begins with the reciprocal character of harmful effects: preventing one activity can impose costs on another. With defined rights and costless transactions, bargaining can move resources towards a more valuable use. In the world of positive transaction costs, however, legal rules and institutional arrangements affect which agreements are possible. His argument asks us to compare the total consequences of feasible alternatives. It cannot be reduced to a promise that private bargaining always resolves pollution. 4
The workshop and its neighbours might agree on containment or a change of process. They can use knowledge that would be costly for a distant authority to acquire. This method works only when the parties can identify the effects and negotiate at reasonable cost. A mutually acceptable agreement avoids some of the delay and expense of imposing a settlement from outside. It also gives those who must live with the result a practical role in shaping it.
Now widen the problem. The residue travels through several neighbourhoods. Its effects emerge slowly and combine with releases from other workshops. Individual losses are small, yet their total is substantial. Finding the sources and organising everyone affected may cost more than any one household could recover. Future residents have no representative at today's negotiation. Under these conditions, an invitation to bargain leaves much of the problem untouched. Collective monitoring and enforceable common rules may make action possible where isolated claimants cannot proceed.
Intentions offer no reliable shortcut. A workshop owner can knowingly avoid a cost. The owner can also follow every assigned responsibility while lacking information about effects further away. Dividing work among departments or businesses creates boundaries in attention as well as in ownership. Asking each person to behave better will help where misconduct causes the damage. It cannot reveal interactions omitted from the information available to a conscientious decision maker.
Physical dispersal increases the difficulty. A contained residue can often be measured and attributed relatively cheaply. Once mixed through soil or water, the same material may require extensive investigation and costly collection. Declaring a property right does not perform that investigation. Rights over diffuse resources can exist, but making them effective demands institutions and measurement. Those costs belong to the comparison of remedies. Treating the resource as ungovernable conceals possibilities; treating governance as costless conceals the work.
This suggests a practical advantage in identifying obligations early. A workshop knows the quantity of a chemical entering its process more readily than neighbours can reconstruct that quantity after residues have dispersed. Recording the relevant use at that stage can reduce the investigation required later. The assessment must still distinguish what the process retains, transforms and releases. Moving observation upstream does not settle responsibility automatically; it can make the information needed to assign responsibility cheaper to obtain. Better measurement can therefore enlarge the range of effects that bargaining or public rules can address.
Payment creates another distinction. A tax can reduce damage by changing behaviour even when its proceeds enter a general budget. Conversely, paying compensation can leave the damaged capacity unrestored. If preservation is our objective, we need evidence about what the response achieves. A transfer of money and a completed repair are separate events. An agreement can be acceptable to its signatories while the productive resource continues to deteriorate beyond their boundary.
I therefore propose starting from the capital an activity uses. Its account would recognise the effort required to maintain that productive potential and the contributions actually made towards doing so. This would make preservation an ordinary obligation in comparing production methods. It would also expose the competitive advantage of a business permitted to shift maintenance costs elsewhere. The method requires decisions about boundaries and acceptable substitution, yet it gives those decisions a definite purpose: determining whether private earnings accompany a maintained productive inheritance or a loss assigned to others.
Braess's extra road returns us to the difficulty: individually sensible choices can worsen shared conditions, so an economic account must follow the costs each decision sends through the system.
References
1. Braess, D., Nagurney, A., & Wakolbinger, T. (2005). On a paradox of traffic planning. Transportation Science, 39(4), 446–450. https://doi.org/10.1287/trsc.1050.0127
2. Lascols, B. (2026). Capital and entropy: Theory of negentropic value (4th ed.). Accounting for externalities.
3. Pigou, A. C. (1932). The economics of welfare (4th ed.). Macmillan and Co. https://www.econlib.org/library/NPDBooks/Pigou/pgEW.html (Original work published 1920)
4. Coase, R. H. (1960). The problem of social cost. The Journal of Law and Economics, 3, 1–44. https://doi.org/10.1086/466560
