all guides
philosophy·17 min read

there are no externalities

depreciation confesses it, deferred maintenance delivers it, and “outside” turns out to be a ledger edge rather than a place

Every accountant knows depreciation. Every facilities manager knows deferred maintenance. Every economist knows externalities.

Three desks, one phenomenon. One confesses it. One delivers it. One gives it an address that does not exist.

Ecological economists have been saying a version of this since 1950. It still has not changed how anything gets funded — which is the part worth arguing about.

the confession

The popular version of depreciation is wrong in a way that matters. Book and tax depreciation do not measure declining market value; they are cost allocation, a rule for spreading money you already spent across the years an asset serves you. A well-maintained building can be fully depreciated on paper while its market value doubles. It is bookkeeping about your basis, not a reading on the asset's condition.

The code lets an owner recover a building's cost over 39 years whether or not a dollar of it goes back into the building. Nothing in the deduction asks about the roof, and two mechanics make the timing aggressive.

Cost segregation. The building shell does not qualify for bonus depreciation: 39-year nonresidential and 27.5-year residential structures are excluded, because §168(k) requires a recovery period of 20 years or less. But an engineered study reclassifies part of the purchase price into shorter-lived classes that do qualify: 5- and 7-year personal property, 15-year land improvements, qualified improvement property. The One Big Beautiful Bill Act made 100% bonus depreciation permanent for qualifying property placed in service after January 19, 2025. Not the whole building — but far faster than any building wears out.

Recapture, and how it disappears. The deduction is a loan against your own basis. Sell, and much of it comes back: real property returns as unrecaptured §1250 gain, taxed at up to 25%, and personal property recaptures at ordinary rates. What the code grants is timing. Two exits turn "later" into "never" — a §1031 like-kind exchange rolls the liability into the next property indefinitely, and a §1014 step-up in basis at death erases it. Defer long enough and deferral becomes forgiveness.

That is the confession, line-itemed on every return: we are recovering capital from an asset, on a schedule that owes nothing to the asset's condition.

the mechanism

Deferred maintenance is how the cost gets delivered — forward, and to someone else.

Facilities managers work from a rule of thumb: a dollar deferred becomes roughly four dollars of capital renewal later. Treat the multiple as folklore with a documented mechanism under it — a leaking roof soaks insulation, wet insulation strains HVAC, strained HVAC fails early. One deferred item manufactures the next.

The standard is not folklore: the National Research Council's Building Research Board concluded in 1990 that routine maintenance plus capital renewal requires 2–4% of a facility's current replacement value, every year, and more than that until an accumulated backlog clears. Almost nobody hits it.

$370B
federal building repair backlog (FY2024)
$181B
Department of Defense facilities
$24B
National Park Service

The federal figure more than doubled from $171 billion between FY2017 and FY2024 — enough that the Government Accountability Office added building condition to its High-Risk List in 2025. Across all US public infrastructure the estimate is about $1 trillion.

None of it is secret: GASB Statement 34 put infrastructure on state and local balance sheets in 1999, making the numbers public and comparable. Disclosure did not produce funding; roughly twenty states publish no total at all.

This is not a market failure but a political economy feature: new construction earns ribbon cuttings on a two-to-six-year electoral clock, maintenance is invisible until something breaks, and the assets last 20 to 100 years. The costs compound and migrate — forward in time, downward in the income ladder, and outward to communities that had no vote in the deferral.

Remember that third direction.

the alibi

Now externalities — with fairness to Pigou first.

Arthur Cecil Pigou built the modern apparatus in The Economics of Welfare (1920): private cost diverges from social cost, and the state closes the gap with a tax. Pigou wanted the producer charged — the opposite of an alibi — and that lineage runs through polluter-pays, Superfund liability, and every carbon price in existence. The concept was not invented to let firms off the hook.

The problem is the noun. Externality comes from Latin externus — "outside, outward, foreign." It describes a relationship, a cost falling on someone outside the transaction, in the vocabulary of a place. Only the first is true. Say "external" often enough and the mind supplies a somewhere-else for the cost to be in.

Karl William Kapp made the structural case in The Social Costs of Private Enterprise (1950): social costs are systematic outputs of how production is organized, not anomalies awaiting a patch.

When most economic activity produces unpriced effects, the externality is not the exception. The priced transaction is the exception.

Scale backs him. In 1997 Robert Costanza and colleagues valued the world's ecosystem services at $33 trillion a year in 1995 dollars — roughly 1.8 times global GNP at the time — while being careful to say most, not all, of that value sits outside markets. Their 2014 update, on a far larger valuation database, put it at $125–145 trillion per year. Argue with the method and you will have company; the authors called the first pass "admittedly crude." You still cannot call something peripheral when it is the substrate the priced economy runs on.

the boundary is chosen

What physics establishes: the residue is permanent. Matter and energy are conserved — the carbon, nitrogen, and sediment leaving a site do not exit existence — and quality degrades, every transformation moving matter-energy from concentrated toward dispersed. Nicholas Georgescu-Roegen framed the economic process exactly that way: low entropy in, high entropy out. Be precise, because the sloppy version gets it backwards: Earth is not a closed box, and the argument does not need it to be. The planet runs a through-flow — solar energy in, degraded heat radiated to space — which is how ordered living systems are possible at all. What Earth does not export is matter. Matter cycles; it does not leave.

What physics does not establish: who owes for it. Cost is not a conserved quantity. No thermodynamic law assigns a dollar of damage to a balance sheet, and no economist ever claimed the physical waste left the planet. Any argument of the form "thermodynamics proves externalities do not exist" is arguing with nobody.

So keep the true version. Externality names a real thing: a cost falling on parties outside the decision that produced it. The quarrel is not with the phenomenon but with the geography the word invents — incidence is not a location, and a ledger boundary is not a wall. The economy is a subsystem of the biosphere, so costs pushed out of a transaction go into the thing the system is inside of: the atmosphere, the aquifer, the soil, the next decade, the town downhill. Nested, not adjacent. The only outside anyone reaches is the margin of their own ledger.

the three concepts, unified

conceptwhat it doeswhat it admitswhat it obscures
depreciationrecovers an asset's cost on a fixed schedulecapital is being used upthat the schedule never asks about condition
deferred maintenancemoves upkeep cost into the futurewe know what is needed and are not doing itwho is standing there when it arrives
externalitynames a cost that lands outside the dealsomething is being harmedthat "outside" is a ledger edge, not a place

An investor who depreciates a building while deferring its maintenance is doing all three at once.

the land exception

One iron rule: land cannot be depreciated. The stated reason, in IRS Publication 946, is that land has no determinable useful life.

The rule is narrower than it looks, and "the tax system is blind to nature" is not true. The code sees nature clearly when nature is being removed: depletion allowances (§§611–613) let owners deduct the exhaustion of timber, minerals, oil, and gas — explicit recognition that a natural stock is being drawn down. Land improvements depreciate over 15 years, and §170(h) grants a deduction for development rights extinguished by a conservation easement.

What the code has no instrument for is condition — the ongoing ecological state of a place, moving up or down, year over year:

  • The building is depreciable — cost recovery runs whether or not it is maintained
  • The land is not depreciable — assumed permanent while its condition moves
  • The ecosystem appears only when something is extracted from it or given away

A forest can lose half its function to beetle kill and a falling water table with no accounting event anywhere.

One correction: classical economics did not treat nature as inexhaustible. Ricardo's "original and indestructible powers of the soil" meant non-augmentable productive capacity; his theory of rent assumes land differs in quality. John Stuart Mill fits the accusation worse — he wrote the founding argument for a stationary state in which growth stops while improvement continues, and defended wild places explicitly. The blindness came later, from national accounting, not from the classics.

So the land exception is the system telling a truth by accident: some things should not be consumed on a schedule. The error is reading a rule about cost recovery as a statement about reality.

what living systems actually do

Depreciation borrows its logic from manufactured capital, where things wear out on a knowable schedule toward salvage value. Living systems run the other way, at least for a long while. A recovering forest appreciates — accumulating biomass, building soil, storing carbon, holding and releasing water. Both words share the root pretium: de- is down, ad- is toward. One direction ends at salvage value and says extract, write off, replace; the other compounds function while processes hold and says protect, maintain, let it work.

Two qualifications a field ecologist would supply anyway. Mature does not mean infinite: net productivity declines as stands age and old systems trend toward a high-value steady state — the compounding is in function and resilience, not a rising biomass number. Disturbance is function, not loss: a grassland burning on its historical interval is a system working, not failing.

This is not metaphor. Ilya Prigogine won the 1977 Nobel Prize in Chemistry for describing such systems as dissipative structures: they hold internal order precisely because they run energy through themselves and export entropy. Applying a salvage-value schedule to that is a category error.

the accounting already exists

Partha Dasgupta's argument is titled "Account for depreciation of natural capital." He is right, and it does not conflict with the section above: two senses of the word are in play. Nature should not be treated as a consumable running down a depreciation schedule — that is the category error. Nature's degradation should absolutely be recorded as depreciation in the national accounts — that is how an invisible loss becomes visible.

And the ledger has largely been built. In March 2021 the UN Statistical Commission adopted SEEA Ecosystem Accounting as an international statistical standard, and its monetary asset accounts explicitly record degradation and enhancement. The World Bank's Changing Wealth of Nations has tracked natural capital within national wealth for years.

Which narrows the problem. Measurement exists. Disclosure exists — GASB 34 for public infrastructure, SEEA for ecosystems. What is missing is the layer neither supplies: money that arrives at the place, on a schedule, to do the work.

Disclosure has never once fixed a roof.

the compounding problem

Ecosystem recovery is slower and shallower than the word suggests. Moreno-Mateos and colleagues pooled 3,035 sampling plots worldwide and found ecosystems in recovery running annual deficits against reference conditions of 46–51% in organism abundance, 27–33% in species diversity, and 32–42% in carbon cycling. They named the accumulated shortfall recovery debt — the function missing during all the years a system spends getting back.

Their conclusion cuts against the lazy version of our own argument: restoration and offsetting are "inadequate alternatives to ecosystem protection." Protection first. Restoration is the expensive, uncertain, slow second choice.

That corrects something this series has said too bluntly elsewhere. Protection is maintenance, and it is the highest-return maintenance available. Removing the driver of degradation is frequently the cheapest effective intervention, and natural regeneration often outperforms active planting. We have a commercial interest in funded intervention, which is why we should say this rather than leave it to a critic.

But protection is not free. Somebody pays for the easement, the fire crew, the water right left in the stream, and the foregone development. Ceasing extraction is continuous funding, not an absence of activity. Unfunded protection is a designation on a map — which is why a park system carrying a $24 billion backlog is worth thinking about.

Where thresholds enter, it stops resembling buildings at all:

built infrastructurenatural systems
replacementbuy anotheroften none, at any price
failure shapegradual, scheduledabrupt, threshold-driven
recovery timemonths to yearsdecades to centuries
reversibilityusually, with moneypast some thresholds, not with any amount

the resolution

None of this argues against internalization — that would run the same trick we just accused the word of running. It is the right instinct, and it has worked when designed well. Costa Rica has paid landowners for watershed and forest services since 1997 and reversed national deforestation. New York City funded Catskills watershed protection instead of a filtration plant and is still drinking the result. Nobody here thinks the field was empty before us.

The pattern in both is the same, and it is not "we priced it": money reached the place and paid for upkeep, on a schedule, for a long time. Pricing was sometimes how the money got raised; it was never the thing that did the work. That gap is what ensurance is built for — not another way to price damage, but a way to measure the condition of a system something depends on and pay against it, before the failure that would make funding obvious. Insurance waits for loss and compensates; ensurance funds protection first.

An agent is an onchain account representing a place, people, or a purpose — a watershed, a land trust, a species. Certificates fund its stewardship, and the distinction inside that word matters. A policy funds a titled asset where a legal titleholder is cooperating, with a committed path toward permanence. A line funds stewardship across boundaries where no single titleholder exists or cooperates, so it carries no legal title, no enforceable maintenance obligation, and no guarantee against lapsing. Most of what exists today is lines. Proceeds route value to the agents; stocks and flows states which condition is being paid for.

Nothing there internalizes an externality. It funds a dependency.

who holds the agent

This essay owes an answer to its own third direction. A mechanism that routed value only to whoever benefits from an ecosystem service would send money to the same asset owners, utilities, and corporations already doing fine — the original arrangement with better reporting. Three governance questions decide whether this is different.

Who holds the agent? Held by people who live in the watershed — or a tribe, land trust, or collaborative accountable to them — the money has somewhere legitimate to land. Held by a distant sponsor who found the watershed on a map, it reproduces representational capture with new plumbing.

Who defines maintenance? "Condition" is not self-evident. Whose baseline, whose evidence counts as proof — Indigenous and local stewards frequently hold longer records and better place knowledge than any monitoring contract will generate. Treat condition as purely technical and you will get it wrong without noticing.

Who gets a share, not just a service? The people carrying deferred costs need standing and payment — as stewards paid for work, as holders of the instruments, as parties whose consent is required — not gratitude for hosting someone else's resilience.

These are design constraints, not solved problems. Whether this is protection or extraction with extra steps is decided in who ends up holding what.

the bottom line

Depreciation is the confession that capital is being consumed. Deferred maintenance is the mechanism by which the bill travels. Externality is the word that gives the bill an address.

There is no address. Long-term warming now sits at roughly 1.4°C above the pre-industrial baseline, with 2024 the warmest single year on record at 1.55°C. The aquifer is lower, the soil thinner, the federal building backlog $370 billion. Every one of those is internal, borne by someone, recorded on a ledger somewhere — just never the ledger where the decision was made.

So the question was never whether to internalize. It is narrower: what are you funding maintenance on, and what are you deferring? Your portfolio has an answer, and so does your city, your supply chain, and whatever acreage you are responsible for. Most have never been written down.

$1 of maintenance now, or roughly $4 of capital renewal later — and for the systems that cross thresholds, no amount at all at the far end.

the series

A series on the words we use to avoid funding what matters.

agree? disagree? discuss

have questions?

we'd love to help you understand how ensurance applies to your situation.