Special Report
What It Means to Cost a Climate Policy
Climate policy generates some of the largest numbers in public life, and they are unusually hard to interpret. Two credible studies of the same measure can differ by an order of magnitude without either being dishonest. This report explains why, and what to look at first when a cost figure appears.
What a cost estimate is
An estimate of the cost of a climate policy is the output of an economic model — usually a general-equilibrium or integrated-assessment model — that represents an economy's sectors, energy system and trade, applies a constraint, and compares the resulting path against a counterfactual path without the constraint. The reported "cost" is the difference between two simulated futures, neither of which will occur.
That is not a criticism. There is no alternative: you cannot run a controlled experiment on an economy, and a modelled comparison is the only rigorous way to reason about a policy that has not yet happened. But it does mean the number inherits every assumption in the model, and models embed a great many.
Four choices that drive the answer
The baseline
The comparison path is a forecast of what the economy would have done anyway — growth rates, energy intensity, technology costs, fuel prices, decades out. A baseline that assumes rapid autonomous improvement in energy efficiency will show a small policy cost, because much of the change happens regardless. A pessimistic baseline shows a large one. The baseline is usually stated in an appendix and almost never in the coverage.
The discount rate
Climate policies impose costs now and deliver benefits over a century or more, so the results depend on how future benefits are converted into present terms. The choice is partly technical — what returns capital earns — and partly ethical, concerning how much weight to give people not yet born. Reasonable analysts adopt rates that differ by a few percentage points, and over a hundred-year horizon a few percentage points change the present value of a distant benefit by more than an order of magnitude. Two studies can share every physical assumption and reach opposite recommendations on this parameter alone. It is the first thing to check.
Technological response
Models differ enormously in how readily they let the economy substitute away from a constrained input. A model with limited substitution options finds the constraint expensive; one that allows rapid deployment of alternatives finds it cheap. Since this is precisely what is uncertain, the assumption largely determines the conclusion.
What is counted
Some studies report gross costs; others net out benefits such as reduced local air pollution, which are often large, near-term and much easier to quantify than the climate benefits themselves. Some include revenue recycling — using the proceeds of a carbon price to cut other taxes — which can substantially reduce or in some models eliminate the net cost. Comparing a gross figure from one study with a net figure from another is meaningless, and it happens constantly.
The forecasting record
There is now a long enough history of environmental regulation to check ex-ante cost estimates against realised outcomes, and the pattern is consistent enough to be worth knowing. Retrospective reviews of major air-pollution and ozone-protection programmes have generally found that actual compliance costs came in below the pre-implementation projections, often substantially. The usual explanation is that models cannot anticipate the innovation a constraint provokes: once compliance has a price, firms discover cheaper routes than anyone had modelled.
Two cautions. This pattern is a tendency, not a law, and it has been found for programmes narrower and more technologically tractable than economy-wide decarbonisation. And ex-ante benefit estimates have their own record, which is more mixed. The honest summary is that pre-implementation cost figures in this field have historically skewed high, and that this should temper confidence in large projected costs without supporting confidence in any particular smaller number.
Reading a study in five minutes
- Find the discount rate. Usually one line in the methods section. If it is not stated, the study is not ready to be quoted.
- Find the baseline. What growth and technology path is the policy being compared against?
- Gross or net? Are co-benefits and revenue recycling included?
- Over what horizon, and for whom? A cumulative figure over thirty years across an entire economy is not comparable to an annual household figure, and both get reported the same way.
- Read the sensitivity table. Most studies publish results under alternative assumptions. That table shows the true width of the answer, and it is nearly always more informative than the abstract.
What is genuinely contested
Setting aside the physical science, which is outside this site's scope, the live economic disagreements are specific and worth naming: the appropriate discount rate for very long horizons; how to value low-probability catastrophic outcomes, which conventional expected-value methods handle poorly; how quickly technology costs fall in response to deployment; and whether policies are best compared on cost per tonne abated or on total abatement achieved. These are unresolved questions among people who agree on the underlying physical facts, and coverage that presents any of them as settled is compressing a real debate.
The Environmental Protection Agency's economics programme publishes the guidelines used for United States regulatory analysis, including its treatment of discounting; the OECD maintains comparative work on the instruments in use across countries; and the Congressional Budget Office publishes accessible summaries of the fiscal side. For the underlying framework, see environmental economics.