The cost of doing nothing (New York)

Every siting refusal has a price. Nobody itemises it.

Refusal is also a decision

An objection to a solar farm is a document of remarkable precision. It counts acres. It surveys the birds. It maps the wetlands, measures the distance to the nearest historic district, and works out the percentage of prime farmland to one decimal place. About the costs of building, objections are never vague.

About the costs of not building, they say nothing at all. The option of refusal does not appear in the ledger even as a zero. A zero would at least be a claim — a figure somebody had calculated and could be asked to defend. What the ledger actually contains is a blank: no information at all. And a blank cell, in a siting decision as in any other accounting, is silently read as zero. Nobody ever decides that doing nothing is free; the assumption simply goes unexamined, because the status quo spreads its costs across people who will never connect them to a hearing room, and hides them from every resident who showed up to speak.

This article puts numbers in that blank.

New York is a useful place to do the arithmetic because, in several respects that matter for this method, it resembles the UK more closely than most American grids do: a broadly comparable latitude and cloud cover, a fossil margin dominated by gas rather than coal, and a level of contest over solar siting that will be familiar to anyone following the UK's own planning fights.

A caution, though: New York's grid is one of the cleanest in the United States, not a representative one. The numbers below are specific to New York; a version of this article run on a more typical, or dirtier, state grid would return a worse result (often far worse) throughout. This is examined in more detail at the end.

A further note on venue. New York created a state authority — the Office of Renewable Energy Siting (ORES) — in 2020, specifically because local refusals of large-scale solar and wind had become routine enough that the legislature intervened. Since then, an outright local refusal — a local body says no, the demand goes unmet — has become rare in New York: most contested projects that would once have gone to a county vote now proceed to a state-level review instead, and get approved, delayed, or modified rather than rejected outright. This article's method still applies to any of those outcomes — the electricity not yet delivered still has to come from somewhere in the meantime — but "the county said no and that was that" is no longer the typical New York story, if it ever was.

The unit of account throughout is a solar farm of sixty-five megawatts — the same nameplate capacity as this article's UK companion piece, chosen for direct comparability rather than because it marks any particular threshold. It sits comfortably inside the range of real New York proposals, which run from small community-solar arrays up through installations several times this size. Rounding throughout is kept against the argument where it matters.

Utility-scale solar in New York operates at a capacity factor of around 15 per cent — measured from NYISO's own generation data for the state's operating solar fleet. This is markedly lower than the roughly 24 per cent achieved by the sunnier, tracker-equipped installations typical of the Southwest, and closer to the UK's own 11 per cent than to the US national average: New York is cloudy, high-latitude, and its solar fleet does not use the loose, single-axis-tracking layouts that push capacity factors up elsewhere in the country. At that rate, a sixty-five-megawatt farm delivers around 85 gigawatt-hours of electricity in its first year — the annual consumption of about eight thousand homes. Panels age, so the accounting assumes output declining by 0.7 per cent a year over a thirty-five-year operating life, the standard degradation rate used in federal technology-cost benchmarks. Over that lifetime, the farm produces about 2.7 terawatt-hours — close to the delivery of the UK companion piece's own farm, which is what shared nameplate capacity and broadly similar weather should produce: Upstate New York and Britain sit at comparable latitudes, and both are cloudy enough that their solar capacity factors land within a few points of each other.

The refused electricity gets generated anyway

Refusing, delaying, or downsizing the application does not make demand disappear. Thousands of homes still keep their well pumps running, and on the New York grid it is possible to say with some confidence what generates the power when they do. Upstate New York's electricity mix is unusually clean for the United States: roughly a third nuclear, a third hydroelectric, and about a quarter natural gas, with coal absent from the state's generation entirely. Nuclear and hydro run at essentially fixed output and cannot be turned up to meet a shortfall; solar and wind generate when the resource permits, regardless of siting decisions elsewhere. The plant that responds at the margin, in the great majority of hours, is a gas-fired unit.

This is not a claim that a megawatt of solar capacity replaces a megawatt of gas capacity. The 85 gigawatt-hours a delayed or downsized farm would have delivered arrives in intermittent blocks, not on demand, and the gas fleet provides the flexibility solar alone cannot. But every one of those gigawatt-hours, whenever it arrives, is one a gas unit somewhere does not have to generate. The fleet stays; it simply burns less.

So the true content of a refusal, delay, or reduction is this: a commitment to burn the natural-gas equivalent of some 85 gigawatt-hours of electricity a year, for what would have been the farm's operating life. The public comment does not phrase it that way. It should. What follows is what that gas costs.

The direct cost is in deaths and serious illness

Burning natural gas produces nitrogen oxides and fine particulate matter, whose effects on the heart and lungs are settled medicine, documented for decades and accepted by every relevant medical body, in the United States as anywhere else.

Rather than reconstruct a mortality-per-terawatt-hour rate from a single epidemiological study, this article uses the same two federal datasets a county or state agency would use if it ran this calculation itself. The EPA's Emissions & Generation Resource Integrated Database (eGRID) publishes a "non-baseload output emission rate" for each region of the country — the emissions rate of the marginal, flexible generating fleet, as distinct from the baseload plants that run regardless. For Upstate New York, that rate is 0.103 pounds of sulfur dioxide and 0.4 pounds of nitrogen oxides per megawatt-hour. The EPA separately publishes, and periodically updates, a benefit-per-ton value for reducing exactly these pollutants from electricity-generating units specifically — $57,000 per ton of SO2 and $7,710 per ton of NOx, in the most recent (2025) estimate, reflecting the monetised public-health cost of a ton of each pollutant emitted from a power plant.

Multiplying these together gives a monetised health cost of about $4.50 per megawatt-hour for New York's actual marginal generating mix — no assumption about the mix's composition required, since the non-baseload rate already reflects it. Applied to the farm's 2.7-terawatt-hour lifetime output, that is just under $12 million in monetised health cost. Converting that to a statistical death count depends on which vintage of the federal government's value of a statistical life (VSL) is used — a parameter federal agencies have applied to regulatory cost-benefit analysis for decades, currently ranging from about $9.8 million (the figure built into EPA's own COBRA health-impact screening tool) to $13–14 million (the higher current estimates used by the Department of Transportation and Department of Health and Human Services). At either end of that range, the result is about one statistical death over the farm's thirty-five-year life.

Illness follows the same pattern. A study using New York State's own hospitalization records found significantly higher rates of hospitalisation for asthma, acute respiratory infection, and COPD among people living near fuel-fired power plants — gas included, not only coal — 11, 15, and 17 per cent higher respectively. Those people are not the applicant, the objector, or the siting board. Who they are is the subject of the next section.

The bill is paid by people who were never asked

It is perfectly reasonable for a community to weigh costs and benefits as it sees them, and decide accordingly. The trouble is that it does not get to see all the costs. A siting review only ever puts the local ones in front of it — the view, the field, the temporary disruption of construction.

The health cost of the previous section falls somewhere else entirely: on the people who live near the gas-fired plants that would otherwise run less. This is not merely a geographic inference. An EPA-adjacent analysis of the nation's "peaker plants" — the flexible gas-fired units that run precisely when marginal demand needs meeting, the same role New York's fossil margin plays in this article's arithmetic — found that nearly two-thirds are sited in communities with an above-average share of low-income residents. A separate, causally-designed study of coal-plant retirements (using the plant closures themselves as a natural experiment) found that reductions in fossil generation produce measurable, disproportionate mortality benefits in the surrounding Medicare population — evidence that the exposure this article is describing is not merely theoretical. The pattern holds even where, as in New York, the specific fuel at issue is gas rather than coal: the people living nearest the plants that run at the margin are not a random cross-section of the state, and they were not consulted on this project.

The public comment and the hospital admission are entries in the same account. They are simply held by different people, and only one of the two knows the account exists.

Accepting mainstream climate science multiplies the bill

Everything so far stands on conventional medicine alone. What follows adds a further cost, for those who accept mainstream climate science; it does not revise what came before. If you reject the climate science, the bill from the previous sections still stands in full — this section only makes it larger.

The 85 gigawatt-hours a year this article's farm would displace carries a second cost. Using the same eGRID non-baseload methodology as the health-cost calculation above — the marginal generating fleet's actual carbon intensity, not an assumption about it — Upstate New York's margin emits about 412 grams of CO2 per kilowatt-hour. Over the farm's 2.7-terawatt-hour lifetime, that is roughly 1.1 million tonnes of carbon dioxide.

Carbon dioxide's death toll is harder to count than a hospital admission's, but it has been counted. The most widely cited estimate — Bressler's "mortality cost of carbon" (Nature Communications, 2021), built on the same modelling that underpins official carbon pricing — puts one excess death this century, worldwide, for every 4,434 tonnes of carbon dioxide emitted. One division later: 1.1 million tonnes is, at the central estimate, on the order of two hundred and fifty statistical deaths — some two hundred times the direct air-quality toll.

That number needs its caveats stated just as prominently. It is a central estimate with wide error bars, from a family of models with well-known limitations; it depends on future emissions, which policy may still change; and the deaths it counts fall across the whole world and the whole century, not on any single community a New York siting board will ever hear from. Reasonable people can halve it or double it. But no reading of the literature supports entering it as a blank.

The ledger, side by side

Here is the complete ledger for a sixty-five-megawatt application — the left column as objections actually itemise it, the right column as this article has now priced it.

The middle column is not a typographical error. It is the objection's actual accounting, reproduced faithfully: it contains nothing, and every siting decision that weighs the left column against it is reading that nothing as a zero. The point is not that the left column is illegitimate — several of its entries are real, and deserve the scrutiny they get. The point is that a comparison between a filled column and a blank one is not a comparison. It is a conclusion reached in advance.

The missing column stays missing because its victims have no faces

Why does the blank persist? Not, for the most part, through bad faith. The economist Thomas Schelling identified the mechanism in 1968: human sympathy, and the institutions built on it, respond powerfully to an identifiable victim and hardly at all to a statistical one — even when the statistical victims are more numerous and just as dead. A named individual in danger can reorganise a community; a change in a mortality rate reorganises nothing. Half a century of research since has confirmed how deep the asymmetry runs.

A siting hearing is that asymmetry, institutionalised. It runs on harms a camera can capture: the field has a location, the habitat has a survey, the view has a viewpoint, and each has a champion with a name and address on the public-comment record. The statistical victims of refusal or delay never enter the process — not because anyone deliberately excludes them, but because no one thinks to include them, and even if someone did, no one could identify them, even in principle. No resident downwind of a gas plant can establish that their admission was one the displaced generation caused; their own doctor cannot tell them; the connection exists only at the level of the population, and populations do not submit public comments. The system is not weighing their interests badly. It has no way for their interests to appear at all.

This is where the pattern this blog has called sudden-onset environmentalism finds its institutional home. The selective precision — exquisite care over hypothetical local harms, silence over real systemic ones — is not just a rhetorical habit of individual objectors. It is the grain of the process itself: every form, incentive, and comment period runs in the direction of the visible column. The objectors did not create the blank. They inherited it, and it flatters their case, so it goes unremarked.

Honest accounting is not the same as automatic approval

Nothing here argues that every solar application deserves consent. Some sites are genuinely wrong — wrong land, wrong place, wrong scheme — and "Solar Spreadsheets" has already made the case that the build side of the ledger deserves honest entries rather than boosterish ones. A ledger doctored in either direction is the same offence.

The argument here is narrower, and harder to escape: a decision made with one column blank is not a decision. It is an accounting error with a mortality rate. A siting board that refuses, delays, or substantially downsizes a sixty-five-megawatt application may, on a full ledger, still be right to do so — but it should do so knowing that the decision is itself a purchase, and knowing the price: on the order of one statistical death and a documented, if not separately monetised, toll of serious respiratory illness on conventional medicine alone; some two hundred and fifty deaths on mainstream climate science; all of it billed to people who never saw the hearing notice.

None of this asks a single board to carry the weight of the energy transition, either. A board prices a single entry; it is the regional ledger — the one "Solar Spreadsheets" argued for — where the entries accumulate and the accounting becomes unavoidable. A state that refuses or indefinitely delays ten such applications has not made ten separate judgements about ten parcels; it has declined twenty-seven terawatt-hours of lifetime generation, one blank cell at a time.

Appendix: showing the working

The farm. Sixty-five megawatts AC, chosen to match the nameplate capacity of this article's UK companion piece for direct comparability, not because it marks a regulatory threshold. It sits within the real range of contested and approved New York applications, which run from roughly 20 MW to several hundred. At New York's utility-scale solar capacity factor of ~15% (derived from NYISO generation data for the state's operating fleet, cross-checked against an independent industry estimate placing a 20% capacity factor as unusually strong performance for the state): ~85 GWh delivered in year one; household equivalence at 10,500 kWh/year (EIA's commonly-cited average US household consumption figure): ~8,100 homes. Operating life taken as 35 years with output degrading at 0.7%/year (NREL Annual Technology Baseline standard degradation assumption), giving a lifetime average of ~89% of initial output and total delivery of ~2.66 TWh. Land use, for reference and not as the article's unit of account: at Bolinger & Bolinger's (2022) empirically-derived tracking-system density of 4.2 acres per MW-DC (LBNL, 736-plant sample, 2007–2019 vintage), a 65 MW-AC farm at a typical 1.34 DC:AC inverter loading ratio occupies roughly 360 acres — consistent with SEIA's commonly-cited fleet-wide range of 5–7 acres/MW.

The counterfactual. Upstate New York (eGRID subregion NYUP), 2023 data: coal 0% (New York's last coal-fired plant closed in 2020); nuclear 32%, hydro 33%, gas 26%, wind/solar/biomass the remainder. Displaced generation is modelled using eGRID's own "non-baseload output emission rate" — the rate EPA itself publishes specifically to answer "what does displaced or avoided generation look like," rather than an assumption constructed for this article. This is the same rate underlying EPA's Greenhouse Gas Equivalencies Calculator and consistent with AVERT's marginal-emissions methodology.

New York's grid is one of the cleaner ones in the country to run this calculation against, and that needs stating plainly rather than left implicit. Coal is a far dirtier energy source, and supplied approximately 17% of US generation as a whole in 2025 (EIA), up from 16% in 2024 — the first year-over-year increase in some years, driven by rising natural gas prices and demand growth. Several states (West Virginia 86%, Wyoming 71%, Kentucky 69%, Missouri 61%, Indiana 46%, per EPA's own eGRID state-level tables) derive a majority or near-majority of their generation from coal. A reader applying this article's method to a proposed farm in one of those states should typically expect a (far) dirtier counterfactual than the one derived here, on both the health and climate figures below — New York's number is closer to a floor than a ceiling.

Direct health costs — the mechanism. Nitrogen oxides and fine particulate matter (PM2.5) from gas combustion are linked to cardiovascular and respiratory disease and premature death by settled medical consensus — the WHO, the American Lung Association, and EPA's own Integrated Science Assessments for particulate matter and ozone all treat this as established rather than contested science.

Direct health costs — the calculation. eGRID2023 non-baseload output emission rate, Upstate NY (NYUP) subregion: 0.103 lb SO2/MWh, 0.4 lb NOx/MWh (annual). EPA Benefit-Per-Ton estimates for electricity generating units, 2025 valuation year (2016$, 2% discount rate): $57,000/ton SO2, $7,710/ton NOx (PM2.5-related). Combined: ~$4.48/MWh in monetised health cost for New York's actual current marginal generation mix. Applied to 2.66 TWh lifetime delivery: ~$11.9 million. Converted to statistical deaths using a range of current federal VSL figures — COBRA's own operational parameter (~$9.8 million, 2016-vintage, inflation-adjusted) and the higher current DOT/HHS central estimates (~$13–14 million) — yields approximately 0.8 to 1.2 statistical deaths, stated in the body as "about one." Caveat: this figure excludes directly-emitted PM2.5, which eGRID does not report a non-baseload rate for; including it would raise the estimate somewhat, making this figure conservative rather than inflated. It also does not separately monetise illness, since the EPA benefit-per-ton values used here are a single figure incorporating both mortality and morbidity effects rather than a Lancet-style breakdown by health outcome; a fuller accounting would disaggregate this into its own dollar or death-equivalent figure, which this calculation does not attempt. The "serious illness" line in the ledger above is left unquantified in dollar terms for that reason — not because no evidence exists for it, as the next entry shows.

Illness — the evidence. Liu, Lessner & Carpenter, "Association between residential proximity to fuel-fired power plants and hospitalization rate for respiratory diseases" (Environmental Health Perspectives 120(6), 2012), used New York State hospitalization records from 1993–2008 and found that living in a ZIP code containing a fuel-fired power plant — a category that includes gas, not only coal — was associated with statistically significant increases in hospitalisation for asthma (11%), acute respiratory infection (15%), and COPD (17%), among people over ten, after adjusting for age, sex, race, income, and rural/urban residence. The same study found no significant association for children under ten specifically, for asthma or ARI. This is cited as direct, New York-specific evidence that the "serious illness" line carries real, non-zero, and non-trivial magnitude — not as a substitute for the formal quantified estimate this article does not attempt, for the reasons given above.

Climate mortality. Upstate NY (NYUP) non-baseload CO2 output rate: 909.3 lb CO2/MWh (eGRID2023), converted to ~412 g/kWh at the standard imperial-to-metric conversion. (The equivalent US national non-baseload rate is 1,372.5 lb/MWh, ~623 g/kWh — half again as high, consistent with the counterfactual note above.) Applied to 2.66 TWh lifetime delivery: ~1.1 million tonnes CO2. Bressler, "The mortality cost of carbon," Nature Communications 12 (2021): one excess death globally, 2020–2100, per 4,434 tonnes (central estimate, baseline emissions scenario; wide uncertainty bounds discussed in the paper). Yield: ~247 statistical deaths, quoted in the body as "some two hundred and fifty." This figure counts combustion CO2 only and excludes upstream methane leakage from natural gas extraction, processing, and transport, which recent literature estimates adds a further burden on the order of tens of percent on top of direct combustion CO2; counting it would increase this figure, not reduce it.

Environmental justice — the evidence base. Clean Energy Group's analysis of EPA power-plant emissions data found that nearly two-thirds of the nation's gas- and oil-fired "peaker" plants — the flexible units that run at exactly the marginal role this article's counterfactual describes — are sited in communities with an above-average share of low-income residents. Fan & Wang, "The impact of PM2.5 on mortality in older adults: evidence from retirement of coal-fired power plants in the United States" (Environmental Health, 2020) uses actual plant retirements as a natural experiment (difference-in-differences design) and finds a measurable, causal reduction in Medicare-population mortality following nearby coal-plant closures — cited here as evidence that the exposure pathway this article describes for gas is a well-established mechanism for fossil generation generally, not a claim that New York's specific counterfactual involves coal.

Pre-empted rebuttal: intermittency. This accounting does not claim that a megawatt of solar retires a megawatt of gas capacity. The 85 GWh figure is delivered energy; the 15 per cent capacity factor has already paid the intermittency bill. Each solar gigawatt-hour, when generated, is a gigawatt-hour of gas not burned. The gas fleet continues to exist and to provide flexibility; it simply runs less.

Pre-empted rebuttal: "build it somewhere else, or on already-developed land." The proposed farm is the option actually on the table, with an application, a financing structure, and an interconnection study behind it. A hypothetical alternative site — a rooftop, a parking lot, a brownfield — typically has none of these, and the argument is rarely advanced by anyone who goes on to champion that alternative once the farm in front of them is refused. An alternative that materialises only as an argument against the real proposal is not an alternative; it is the blank column with better production values.

Pre-empted rebuttal: "building it has a carbon cost too." True, but small and already favourable, on the same logic as the UK companion piece: utility-scale solar PV pays back its embodied carbon in one to three years against a 35-year operating life, and the comparison is in any case already asymmetric in the objector's favour, since it costs no embodied carbon for the displaced gas plant's own construction and fuel-supply infrastructure.

Pre-empted rebuttal: "it will simply be built somewhere else." New York created ORES in 2020 specifically because local refusals had become routine enough to threaten the state's own legislated clean-energy targets — a fact that undercuts, rather than supports, the claim that any given refusal is costless because the capacity lands elsewhere. If displacement were frictionless, the state would not have needed to override local authority to hit its own targets. This does not mean every application should be approved; it means the "it'll happen anyway" defence is inconsistent with the reason the current siting process exists in the first place.

Drafted in collaboration with Claude (Anthropic).