The Nuclear Counterfactual

A model-dependent ledger of foregone clean power

Between 1965 and 1975 the United States rapidly expanded nuclear construction. The subsequent slowdown involved public opposition and changing regulation, but also cost overruns, interest rates, electricity-demand forecasts, utility finance, project management, accidents, and competition from other generation. France’s more standardized program is an informative comparison, not a blueprint the United States could be assumed to reproduce without institutional differences.

I want to do two things with that turn. First, diagnose it — name the error precisely, because it is a specific and instructive kind of error, not just a bad bet. Second, price it. The diagnosis without a number is a complaint; the number without a diagnosis is a provocation. Together they make a claim you can argue with, which is the only kind worth making.

The Shape of the Mistake

The anti-nuclear turn was fear-driven, and the fear was real but miscalibrated. Three Mile Island released no measurable harm to the public and killed no one; it terrified a nation anyway. The word “nuclear” carried freight the reactors did not — mushroom clouds, invisible poison, waste that lasts ten thousand years. A movement that called itself environmentalism organized around opposing the technology, and that opposition reshaped policy, investment, and public sentiment for half a century.

Germany’s post-Fukushima shutdown is an important case, not a clean natural experiment. Direct radiation effects from Fukushima were far smaller than early public fears, while evacuation and disruption produced serious health costs. Germany simultaneously changed renewables, coal, gas, efficiency, imports, and climate policy; estimates of the shutdown’s emissions and mortality effects depend on the generation counterfactual and time window. The strong claim that retiring low-carbon firm generation increased reliance on fossil alternatives is defensible. Assigning the entire path to panic is not.

This creates a burden of coherence. Anyone advocating rapid decarbonization while opposing nuclear must show that renewables, storage, transmission, demand response, firm low-carbon alternatives, and timing can replace the lost output with acceptable cost and reliability. Nuclear advocates owe the reciprocal showing on construction risk, waste, proliferation, accident liability, financing, and opportunity cost. Climate urgency does not logically make one technology indispensable, but it raises the cost of excluding a proven low-carbon option without a replacement plan.

None of this requires imagining a utopian all-nuclear grid or flawless execution. The counterfactual is modest: suppose the environmental movement had campaigned for clean air, climate stability, and energy independence by championing nuclear power instead of fighting it. Political support in place of political obstruction. Nothing more heroic than that. Then ask what the obstruction cost.

Five Ledgers

The counterfactual can be modeled, but not directly measured. The five ledgers below are an illustrative scenario calculation. A credible estimate would require a reproducible generation model, build schedule, capital costs, displaced fuel mix, epidemiological functions, policy feedback, uncertainty distributions, and checks against double counting. Surface arithmetic is useful for exposing assumptions; it does not make the assumptions validated.

The scenario assumes roughly 125,000 TWh of U.S. fossil generation across five decades and stipulates that a nuclear-first path displaces 30%, or 37,500 TWh. That displacement fraction is the load-bearing counterfactual. It must not be described as observed or as “France’s” without modeling differences in demand, construction, grids, and institutions.

Mortality. Burning fossil fuels is not an abstraction; air pollution contributes to strokes, heart attacks, respiratory disease, and premature death, with rates differing sharply by fuel, pollution controls, location, exposure model, and period. If this scenario assigns an average of 40 avoided premature deaths per displaced TWh, 37,500 TWh yields 1.5 million. To display the monetization, apply a value of a statistical life of $10 million:

\[1.5 \text{ million lives} \times \$10\text{M} \approx \$15 \text{ trillion}\]

This is the largest entry because the selected rate and VSL dominate the arithmetic. Neither is sourced or validated here for this five-decade U.S. fuel mix, so the result is a scenario output, not a conservative empirical estimate.

Carbon. Coal and gas emit roughly 0.7 tons of CO₂ per MWh; nuclear is near zero. Displace a third of fossil generation for fifty years and you avoid about 26 gigatons of CO₂. To price it, the social cost of carbon — the present value of the future damage each ton does — runs anywhere from $50 to $200 per ton depending on the discount rate and the damage function you accept. That range is the output the science supports, not a point estimate:

\[26 \text{ Gt} \times \$50\text{–}200/\text{ton} \approx \$1.3\text{–}5.2 \text{ trillion}\]

Midpoint, about $2.6 trillion — and this captures only the monetized slice, nothing for ecosystem loss, ocean acidification, or climate-driven instability.

Electricity cost. A nuclear-heavy grid front-loads capital and can reduce exposure to fuel-price shocks, while construction cost, financing, overruns, utilization, transmission, and displaced alternatives determine total system cost. If the scenario stipulates annual electricity spending of $450 billion and a sustained 10 percent saving, the arithmetic is:

\[\$450\text{B} \times 10\% \times 50 \text{ years} \approx \$2.25 \text{ trillion}\]

Geopolitics. Energy dependence can affect foreign policy and defense commitments, but a more nuclear electricity system does not directly specify oil dependence, military choices, or their counterfactual costs. The causal chains are too long to price from the assumptions supplied here. The table retains $2 trillion only as an explicit placeholder whose removal is part of the sensitivity test, not as an estimate.

The lost industry. Sustained U.S. buildout could have preserved more reactor construction, fuel-cycle, and engineering capacity, but market share would also depend on execution, cost, diplomacy, regulation, and competitors. The table’s $3 trillion is another placeholder without an industry model, not a measured loss.

The Total, and How to Move It

Add the scenario values:

The table sums to about $25 trillion under the selected inputs. It is not a conservative estimate or a validated confidence interval. The mortality figure scales an assumed displacement through an epidemiological rate and then a regulatory value of statistical life; the electricity, geopolitical, and industrial ledgers require independent models and may overlap with other benefits. A defensible publication should provide sources, units, uncertainty ranges, and a spreadsheet before presenting a total as an empirical result.

Now the part that makes the number trustworthy rather than merely large: it is built to be attacked, and I will show you the handles. Halve the VSL to $5 million — some economists prefer it — and the mortality ledger drops from $15 trillion to $7.5 trillion, taking the total under $18 trillion. Take the low end of the social cost of carbon and the carbon ledger nearly halves. Think the displacement fraction should be 20 percent rather than 30, because France’s grid is not America’s and execution is never clean? Scale every physical quantity down by a third and the whole estimate shrinks with it. Doubt the geopolitics ledger entirely? Strike it; the total barely notices, because it was the smallest and softest to begin with. Push the other way — a higher VSL, a higher carbon price, a larger displacement — and you climb toward the $50 trillion end.

Sensitivity is necessary, but a range produced by varying hand-selected inputs is not yet an estimate. It shows which assumptions dominate and which research would change the conclusion. Fossil combustion causes material health and climate harms; the incremental harm attributable to a foregone nuclear program still depends on what would actually have been built, displaced, and operated.

The Doomer Symmetry

Step back and the shape of the error is familiar. It is the shape Lessons From Peak Oil diagnoses: a catastrophist reasoning failure — a vivid worst case, statically imagined, its probability inflated by dread, its costs of avoidance ignored. Peak-oil doomers fixated on running out of oil and never priced in substitution and adaptation. Anti-nuclear campaigners fixated on meltdowns and never priced in the coal smoke they were choosing instead. Both are catastrophism. The peak-oil version predicted a disaster that adaptation prevented; the nuclear version predicted a disaster so rare it barely registers in the fatality statistics, and in fighting it caused a slower, larger disaster measured in the fossil pollution ledger above.

This is the symmetry worth holding onto: the anti-nuclear movement was itself a doomer movement. It failed the same field test peak oil failed — it took a low-probability tail risk, treated it as the dominant consideration, and discounted the systemic cost of the alternative it was steering toward. The five questions that catch peak-oil catastrophism catch this one too. What is the probability, really, weighed against the base rate? What does the feared outcome actually cost when it occurs, versus the outcome you get by avoiding it? What adapts? What does the price system already handle? Run nuclear fear through that filter and it fails, exactly as the oil scare failed, but with a body count.

And that is what makes this the darkest entry in a volume otherwise about great progress. The story of the modern world is child mortality falling from one in three to under one in a hundred, life expectancy climbing decade on decade, hot showers no pharaoh could command. That progress is real and it is the default, but it is not automatic. It runs on cheap, abundant, reliable energy, and it can be slowed by a society whose conceptual filters amplify fear instead of coherence. The nuclear counterfactual is progress foregone — a half-century where the curve could have risen faster, the air been cleaner, the carbon lower, and did not, because a movement mistook the ticket for the coat and its own fear for a fact.

History rarely hands us a clean energy counterfactual, and this is not one. The durable lesson is to compare complete systems: nuclear construction and accident risks, fossil pollution and climate costs, renewable variability and transmission, storage, demand, and institutional execution. Arithmetic disciplines the story only when the model, sources, uncertainty, and counterfactual are open enough to reproduce.