Dr. Wallace's Favorite Mental Models

Externalities

Every transaction has two people in it and a price that only reflects what those two people value. When the deal also dumps a cost, or hands out a benefit, to someone who was never at the table, the price is lying about the true size of the thing. That gap between private cost and social cost is an externality.

The 30-second version

What it is

An externality is a cost or a benefit that falls on someone who had no say in the decision that created it.

The economist Alfred Marshall noticed in his 1890 Principles of Economics that a firm's output can quietly raise or lower the fortunes of other firms nearby, calling these "external economies." Arthur C. Pigou turned the observation into a policy tool in The Economics of Welfare (1920): if the private cost a firm pays doesn't match the full social cost of what it does, tax the gap for a harm or subsidize the gap for a benefit, and the firm's own self-interest will do the rest. In 1960, Ronald Coase complicated the story in "The Problem of Social Cost," arguing that when property rights are clear and the people involved can negotiate cheaply, they often sort out the externality themselves, no tax required — the puzzle is less "who's wrong" than "who can bargain."

Economists sort externalities two ways: by direction (does it help or hurt bystanders?) and by side of the market (does it come from making something or from using it?). The four combinations:

−

Negative production

A factory's smoke damages its neighbors' health and property. The firm's cost sheet never sees that bill.

+

Positive production

A beekeeper's hives pollinate the orchard next door for free. The orchard's harvest improves and nobody invoiced anybody.

−

Negative consumption

A loud apartment or a chain-smoker's secondhand smoke costs the people nearby, not just the person making the choice.

+

Positive consumption

Getting vaccinated protects you, and it also lowers the odds that the people around you catch the disease from you.

One-line test Before trusting a market price as "the" price of something, ask: "Does this number include every cost and every benefit that lands on someone who isn't part of the deal?"

Mental Model Compass

Where this model sits

The horizontal axis asks whether a cost or benefit stays with the person deciding or spills onto other people. The vertical axis asks whether a model is mainly a design tool you reach for on purpose, or mainly a warning sign that something has already gone wrong. Choose a model to compare it with Externalities.

Mental Model Compass A two-axis map. Left means the cost stays private, right means it spills onto others. Top means a design tool, bottom means a warning sign. Externalities sits centrally, a little toward social spillover and a little toward design tool. Incentives sits top-left. Network Effects sits top-right. Supply and Demand sits left of center at mid-height. Information Asymmetry sits right of center at mid-height. Moral Hazard and Tragedy of the Commons sit toward the bottom right, closer to warning sign. Cost stays private Spills onto others Design tool Warning sign
Build it step by step

From a missing price to a working fix

Step 1 of 5

Private cost and social cost are different numbers

A firm's private cost is whatever it actually pays: labor, materials, fuel. The social cost is the private cost plus every cost the activity imposes on everyone else: the asthma from the smoke, the noise, the traffic. When the two match, the market price is trustworthy. When they don't, the market produces too much of harmful things and too little of helpful ones, because the person deciding only feels their own slice of the bill.

Social cost = private cost + external cost

Step 2 of 5

Spotting the direction

Ask two questions: who is making the decision, and who else feels the result? If the uncounted effect hurts bystanders, the activity will be overproduced relative to what's efficient. If it helps bystanders, the activity will be underproduced, because the decider never gets credit for the upside.

Quick check: a software company under-invests in writing documentation, because the main beneficiaries are other teams and future hires, not this quarter's shipping metrics. What does the model predict?

Step 3 of 5

Coase's surprise: sometimes nobody needs to referee

Ronald Coase's 1960 paper made an argument that startled his own colleagues at a famous University of Chicago dinner where he reportedly won over 20 skeptics by the end of the evening: if property rights are clearly assigned and the parties can negotiate at low cost, they will bargain their way to the efficient outcome regardless of who initially holds the right. A rancher and a farmer sharing a fence line can work out who pays for it without a regulator, as long as talking to each other is cheap.

Quick check: why doesn't the Coase theorem's bargaining solution work well for global carbon emissions?

Step 4 of 5

When bargaining can't scale, price it or cap it

Pigou's answer for large, diffuse externalities: attach a price. A Pigovian tax set equal to the external cost makes the decider feel the full social cost, so they naturally cut back to the efficient amount. A Pigovian subsidy does the mirror image for positive externalities. A third tool, cap-and-trade, fixes the total quantity allowed and lets a market set the price: the 1990 U.S. Acid Rain Program capped sulfur dioxide from power plants and let them trade permits, cutting SO₂ emissions roughly in half by the early 2000s at a fraction of the cost industry had originally projected.

Step 5 of 5

The practical rule

Find the missing price, then pick the cheapest fix that actually scales to the number of people involved. Name who bears the uncounted cost or gets the uncounted benefit. If it's a handful of parties who can talk to each other, let them bargain over clearly assigned rights. If it's thousands or millions of anonymous bystanders, price the externality directly (tax, subsidy or a capped, tradable permit) rather than hoping for a conversation that can never happen.

You'll try the pricing fix yourself in the simulator under "Apply it."

Across domains

Real examples

Climate policy

Pricing a ton of carbon

The U.S. government's own Social Cost of Carbon estimate, the dollar value assigned to the long-run damage from one metric ton of CO₂, was about $51 under the interagency working group's 2021 figure. In 2023 the EPA proposed raising its central estimate to roughly $190 per ton, reflecting updated climate science and economic modeling. Either number is an attempt to put Pigou's idea into practice: make the price of emitting reflect the damage it causes.

Lesson: The "right" external cost is an estimate, not a fact carved in stone, and it gets revised as evidence improves — which is exactly what happened here.

Agriculture · classic case

Bees and the orchard next door

The economist James Meade used beekeeping in a 1952 paper as the textbook case of a positive production externality: a beekeeper raises hives to sell honey, and the bees pollinate neighboring apple and almond orchards for free, raising the orchard's yield. Neither side bills the other, so left alone, the market tends to under-supply hives relative to what would be efficient for both honey and fruit.

Lesson: Modern commercial beekeeping in California's almond belt now runs on paid pollination contracts, a real-world Coasean fix that turns the externality into a priced service.

Urban transport · London, since 2003

The congestion charge

Every car entering central London on a weekday pays a daily charge. Each driver's decision to enter clogs the road for everyone else and adds to local air pollution, costs the driver never saw on their own dashboard. Traffic volumes inside the original charging zone fell by roughly 30% in the scheme's early years, and researchers using the policy as a natural experiment have linked it to measurable local reductions in traffic-related air pollutants.

Lesson: A per-trip price, not a ban, let the people with the most urgent reasons to drive keep driving while discouraging the marginal, low-value trips.

Public health

Herd immunity from a vaccine

Getting vaccinated mostly protects the person who gets the shot, but it also lowers the chance they pass an infection on to someone who can't be vaccinated, an infant, an immunocompromised relative, a stranger on the train. That extra protection to bystanders is a positive externality, and it's the standard economic argument for public subsidy or mandate of vaccination: left to private incentives alone, uptake tends to land below the level that's best for the group.

Lesson: The individual's cost-benefit math and society's cost-benefit math point to different "right" amounts of vaccination, by design.

Knowledge work

The open office and the borrowed attention

A loud phone call, a drive-by question, or a chat notification costs the interrupter almost nothing but can cost a nearby colleague far more. Research by Gloria Mark and colleagues at UC Irvine (presented at the CHI conference in 2008) found that, in an office study, it took people an average of about 23 minutes to fully return to a task after an interruption. The interrupter rarely pays that 23-minute bill.

Lesson: "Quick questions" are a negative externality on focus; norms like scheduled office hours or async-first communication are a Coasean fix that reassigns who has to do the inconveniencing.

Software engineering

The shortcut that bills another team

A team under deadline pressure ships an undocumented, tightly-coupled change to a shared API. It hits their sprint goal at zero extra cost to them. The cost shows up later as debugging time, broken integrations and onboarding confusion for every other team that touches that API, costs the shipping team's own metrics never register.

Lesson: Internal "externalities" like this are why some engineering organizations assign an explicit owner, a review gate, or a charge-back for shared-infrastructure changes: it reattaches the cost to the decision that caused it.

Worked application

Pricing the externality at a power plant

Illustrative scenario Riverbend Generating is a fictional coal-fired power plant with made-up demand and cost curves, built only to show the mechanics clearly. The externality-cost slider below, however, is anchored to real numbers: the U.S. government's own Social Cost of Carbon estimates of roughly $51 per metric ton (2021 interagency working group) to roughly $190 per metric ton (EPA's 2023 proposed update), translated into dollars per megawatt-hour using the standard industry rule of thumb that a coal plant emits approximately one metric ton of CO₂ per MWh generated.

The steps

  1. Find the private equilibrium. Where the plant's own marginal cost meets market demand is where an unregulated plant will produce.
  2. Add the external cost. Social marginal cost is private marginal cost plus the harm per unit, here driven by the carbon-cost slider.
  3. Find the social optimum. Where social marginal cost meets demand is the output level that's actually efficient for everyone.
  4. Measure the gap. The triangle between the private and social outcomes is the deadweight loss: real value destroyed that nobody captures.
  5. Apply a Pigovian tax equal to the external cost. This is the lever that, in theory, moves the market from the private outcome to the social one without a regulator picking a quantity by hand.

What to watch in the simulator

Drag the external cost per MWh slider from $51 toward $190 with the tax switched off: notice the gap between what the plant produces and what's efficient keeps widening, and so does the deadweight loss.

Flip the Pigovian tax switch on at any slider value: the plant's effective cost curve shifts up by exactly the external cost, output falls to the socially efficient level, and the deadweight loss collapses toward zero.

Anchored to real Social Cost of Carbon estimates: $51/ton (2021) to about $190/ton (EPA, 2023), at roughly 1 ton CO₂ per MWh of coal power.
Apply a Pigovian tax equal to the external cost
Misapplications

Common mistakes

Calling every side effect an "externality"

Competition that puts a worse product out of business isn't an externality; it's the market working as intended. Fix: Reserve the word for effects that bypass the price system entirely, not for ordinary winners and losers inside it.

Assuming every externality needs a government fix

Coase's point was that small-numbers situations with clear property rights often sort themselves out through negotiation, cheaper and faster than regulation. Fix: Check the number of affected parties and the cost of talking to each other before reaching for a tax or a law.

Taxing the wrong thing

A tax on a firm's revenue doesn't shrink the harmful activity itself; it just makes the firm poorer. Fix: Price the specific unit that causes the harm (a ton of emissions, a liter of effluent), not a proxy for it.

Treating the dollar value of the externality as settled science

The Social Cost of Carbon moved from about $51 to about $190 per ton between official U.S. estimates as models and evidence changed. Fix: Build in a process to revisit the number, and say out loud that it's a current best estimate, not a constant.

Overcorrecting with a tax set above the true external cost

A tax that's too high creates its own deadweight loss, just in the opposite direction, by discouraging activity that was actually worth doing. Fix: Anchor the tax to the best available damage estimate, not to "as much as the activity can bear."

Forgetting that positive externalities need the opposite medicine

People fixate on pollution and copy the tax tool for every case, including ones where the spillover is good (education, vaccination, open-source software) and the fix is a subsidy or a mandate, not a tax. Fix: Settle the direction first, then pick the matching tool.

Boundary conditions

When not to use it

Ordinary private trade-offs

Deciding whether to buy or rent, or which supplier to use, is a regular cost-benefit decision. There's no third party footing part of the bill, so it isn't an externality question at all.

Effects too small to matter

If measuring and pricing the spillover would cost more than the spillover itself, the model is technically correct but practically useless. Not every externality clears the bar for intervention.

Low-cost bargaining is available

When a handful of clearly identified parties can negotiate directly and cheaply, reaching for heavy regulation first, instead of letting them strike a deal, usually wastes time and goodwill.

The damage estimate is genuinely contested

Forcing a precise tax rate onto a number that serious researchers still disagree about by a wide margin can lock in a wrong price and crowd out the debate that would have improved the estimate.

Pure transfers, not new costs

Money moving from one pocket to another, with no extra resource used up or created, is a distribution question, not an externality. Gambling winnings and inheritance aren't externalities.

Disapproval dressed up as harm

Disliking a neighbor's taste, lifestyle or private choice is not the same as bearing a measurable cost. Economists generally exclude purely moral or aesthetic discomfort from the external-cost ledger, and treating it as equivalent invites the tool to be used as a weapon.

Try it yourself

Build a prompt for your own situation

Fill in the fields. The prompt updates as you type, ready to paste into any AI assistant.

Your prompt


        
        

      

    

  







  

    
Save as skill

    
Take this model with you

    
This Markdown file gives an AI assistant a compact, reusable version of the model. Paste it into custom instructions, a project knowledge file, or anywhere your assistant reads standing guidance.

    

      

        
Skill preview

        

        

          
          
        

        

      

      

        
Apply to my problem

        
A more action-focused prompt: describe a decision or policy you're weighing, and get back a concrete plan for pricing or negotiating the spillover.

        

        

        

        

        
        

      

    

  






  

    
Further reading

    
Sources

    
    
      
  • Marshall, A. (1890). Principles of Economics. Macmillan.
    • 
      
  • Pigou, A. C. (1920). The Economics of Welfare. Macmillan.
    • 
      
  • Coase, R. H. (1960). "The Problem of Social Cost." Journal of Law and Economics, 3, 1–44.
    • 
      
  • Meade, J. E. (1952). "External Economies and Diseconomies in a Competitive Situation." The Economic Journal, 62(245), 54–67.
    • 
      
  • Interagency Working Group on Social Cost of Greenhouse Gases (2021). Technical Support Document, central estimate of approximately $51 per metric ton CO₂ (2020, 3% discount rate).
    • 
      
  • U.S. EPA (2023). Proposed update to Social Cost of Greenhouse Gases, central estimate of approximately $190 per metric ton CO₂ (2020, 2% near-term discount rate).
    • 
      
  • U.S. EPA, Acid Rain Program retrospective reporting on SO₂ emissions reductions under Title IV of the 1990 Clean Air Act Amendments.
    • 
      
  • Mark, G., Gudith, D., & Klocke, U. (2008). "The Cost of Interrupted Work: More Speed and Stress." Proceedings of the ACM CHI Conference on Human Factors in Computing Systems.
    • 
      
  • Transport for London and independent academic studies of the London congestion charging scheme's effects on traffic volume and local air quality since 2003.
    •