Calibrated forecasts of how climate affects economic outcomes.

Price the climate signal. Our forecasts run from one-month to multi-year horizons across operational and strategic windows, published with interpretable uncertainty intervals, out-of-sample validation, and the macroeconomic controls that isolate the climate effects from market dynamics.

Recent research

Climate-economic transmission across commodity, industrial activity, portfolio earnings, and transition risk.

Commodity · Industrial activity · Portfolio earnings · Transition risk

Seasonal and multi-year outlooks, with calibrated intervals, climate mechanism, and out-of-sample validation.

Two seasonal forecasts and a multi-year resilience outlook, each published with its calibrated intervals or probabilities, the climate mechanism driving the signal, and the backtest or simulation validation behind it. Each is also situated against the historical range of normal variation.

PRELIMINARY · SEAS5 ENSEMBLE · PRE-SOWING
Indian Cotton, 2026 Monsoon Season (Kharif)
−5.2%
Forecast yield anomaly

Modal outcome is normal range (61.8% probability) with an asymmetric downside tilt. Surplus risk is low (12.7%). The signal is ENSO-driven across all five clusters. El Niño emergence during the kharif season typically weakens monsoon moisture delivery to the cotton belt.

Risk tercile probabilities
25.5%
61.8%
12.7%
Below Normal Above
Calibrated intervals Bands narrow as each phase completes
−45.4% −29.5% median −5.2% +17.2% +33.8%
Phenological timeline
Preseason
Sowing
Vegetative
Sq./Flower
Boll dev.
Maturation
Coverage
Five regional clusters across 116 districts, seven states. Dominant cluster: Maharashtra, Gujarat, Madhya Pradesh, Andhra Pradesh, Telangana.
Validation Walk-forward IC 0.29 · ICIR 1.76 · 82% national directional accuracy · CRPSS +0.57 Post-Bt era only (2002–2024). NASA POWER reanalysis + ECMWF SEAS5 ensemble. NOAA CPC ENSO data. Forecast updates once June sowing climate arrives. Full validation: Cotton Working Paper →
DEVELOPING · SEAS5 ENSEMBLE · 2 OF 6 FEATURES OBSERVED
US Cotton, 2026 Season (Georgia)
+1.1%
Forecast yield anomaly

Near-normal season with a slight upside lean. 79.0% probability of normal range, 5.2% below-normal, 15.8% above-normal. The distribution is asymmetric: the lower tail runs to roughly −6% in stress scenarios, while the upper tail reaches +14%. Pre-planting dewpoint signals are observed; three growth phases remain ahead before the forecast resolves.

Risk tercile probabilities
5.2%
79.0%
15.8%
Below Normal Above
Calibrated intervals Bands narrow as each phase completes
−6.0% −5.6% median +1.1% +9.7% +14.2%
Phenological timeline
Preplanting
Planting
Vegetative
Flowering
Boll dev.
Coverage
Eighteen counties across Georgia, modelled on USDA NASS yield data 1996–2024. Texas and California outlooks are pre-season; forecasts will surface as their phenological calendars open.
Validation Walk-forward 100% aggregate directional accuracy · IC 0.924 · CRPSS +0.675 · 70% panel directional accuracy at the county level NASA POWER climate reanalysis + ECMWF SEAS5 ensemble. Early-season signal from pre-planting dewpoint variability; flowering heat stress carries the remaining signal weight. Forecast updates as flowering and fruit-setting climate observations arrive. Full validation: Cotton Working Paper →
MULTI-YEAR RESILIENCE · BASELINE + CLIMATE-CHANGE SCENARIO
California Almonds, 20-Year Resilience Outlook

The risk is not a single bad season. It is the compound probability of multi-year disruption and the recovery trajectory that follows.

California almonds face a specific climate-risk profile that seasonal yield forecasts alone do not capture. Bloom frost is the dominant stressor — immediate crop loss with fast recovery — but it operates alongside water allocation shortage, extreme heat episodes, and orchard removal decisions that can convert a bad season into permanent capacity loss. We simulate the compound probability of disruption across a 20-year horizon using 5,000 Monte Carlo runs that propagate each stressor with its historical frequency, severity, and recovery dynamics.

Baseline 20-yr Climate-change 20-yr
P(>10% disruption in any year) 86.9% 89.1%
P(>20% disruption, 2 consecutive yrs) 40.6% 46.6%
P(permanent capacity loss) 21.6% 25.9%
Median maximum disruption 22.2% 24.5%
Mean recovery time 3.1 years 3.5 years
Methodology Resilience simulation: 5,000 Monte Carlo runs × 20 years, stressors calibrated from historical frequency and severity distributions. Four stressors modelled: bloom frost (primary), water allocation shortage, extreme heat wave, orchard removal. Climate-change scenario applies documented shift in stressor probabilities. Baseline yield reference 2,014 kg/ha.

Forecasts are updated as new climate observations arrive within the growing season. Previous issues are preserved on the public record (see methodology below).

Coming East Africa coffee yield forecast, Q2 2026. Almond yield & quality forecast (California, Spain), Q3 2026.

Commodity · Industrial activity · Portfolio earnings · Transition risk

How industrial technologies degrade under specific climate conditions.

We estimate technology-specific dose-response curves — the relationship between a climate variable and a facility's operational efficiency — on multi-thousand-facility panels. The curves are functional-form selected by out-of-sample fit, validated on held-out data, and segmented by production technology rather than by sector aggregate.

01

Electric arc steelmaking and heat stress

WBGT (wet-bulb globe temperature) × production efficiency
Efficiency anomaly +5% 0 −5% −10% −2σ −1σ 0 +1σ +2σ Threshold WBGT deviation from facility-specific baseline

Across 452 electric arc furnace (EAF) steel facilities, we find a non-linear dose-response between wet-bulb globe temperature (WBGT) deviation and production efficiency. Efficiency degradation begins at moderate breach magnitudes and accelerates above the threshold: the cost of an extreme breach is disproportionately larger than the cost of two moderate ones.

The non-linear shape gives a capital-allocation question a quantitative answer. Cooling-system upgrades at EAF facilities are expensive, and their efficiency payback depends on the WBGT distribution the facility actually encounters. The dose-response curve, combined with facility-specific climate projections, turns that payback from a qualitative argument into a calculable expected return.

Validation N = 452 facilities · Period 2021–2025 · Non-linear functional form selected by out-of-sample fit · Climate signal isolated through facility-level controls · Climate TRACE + ERA5. Working paper available on request.
02

Integrated-dry cement and vapour pressure deficit

VPD × production efficiency, with duration-dependence
Efficiency anomaly +5% 0 −5% −10% Low VPD Median High VPD Short breach Extended breach VPD deviation · short-breach vs extended-breach response

Across 2,078 integrated-dry cement facilities, vapour pressure deficit shows a non-linear dose-response with a statistically significant duration-dependence. The cost of a breach extends beyond the breach itself: efficiency does not return to baseline within the period following a breach episode, and the longer a breach persists, the greater the post-breach impact.

Under CMIP6 climate trajectories, VPD is projected to rise materially across several cement-producing regions — particularly in South-East Asia — over the next two decades. The duration-dependence we observe converts those projections into a concrete operating-cost trajectory, with the efficiency penalty compounding across extended breach episodes that are becoming both more frequent and more persistent. The resulting question for an integrated-dry operator is not whether heat exposure matters, but how much operating margin the facility loses when breach duration doubles or triples — a question the curve now makes quantifiable.

Validation N = 2,078 facilities · Period 2021–2025 · Duration-dependence statistically significant at conventional levels · Recovery dynamics under continued review; quantitative recovery estimates available in the working paper.

We hold the rest of the dose-response panel — across heavy industry and consumer sectors and remaining climate variables — under continued methodological review. Working paper coverage of cement, steel, pulp, textiles, and food processing is scheduled for release later this year.

Calibrated at facility resolution, these same threshold features feed the company-level work on margin and earnings developed in the section that follows.

Commodity · Industrial activity · Portfolio earnings · Transition risk

The climate-earnings transmission gap.

Major institutions — NBIM, ECB, BIS — have flagged the gap between observed climate operational impact and corporate earnings disclosures. The transmission mechanism is established econometrically; the empirical magnitude at the firm level remains an open research question.

Our current work focuses on company-level case studies in cement, pulp, and food processing, addressing measurement-error limitations that have constrained prior cross-sectional approaches. Initial results in preparation.

Commodity · Industrial activity · Portfolio earnings · Transition risk

Where financial capacity allows decarbonisation, and where it doesn't.

Where the transmission-gap research maps the operational cost climate is imposing on margins, the Carbon Frontier research applies the same analytical logic to transition risk — a second channel through which climate is mispricing affected equities. Most sustainability analysis tracks whether companies are meeting their stated decarbonisation targets against distant pledges. We ask a different question: what climate performance the financial profile of a company — in particular its leverage, return on capital, and profitability — allows, and where the company sits relative to that frontier. Put another way: we evaluate transition risk before it shows up in sustainability reports, by connecting climate performance directly with economic performance.

Frontier dispersion · 103 listed companies
Each dot is a company. Position shows distance from the financial frontier and direction of travel.
Five labelled cases (A–E) span the range of positions the framework identifies.
FRONTIER +500% +200% +100% +50% +10% 0 −50% +100 +30 +10 0 −10 −30 −100 widening gap from frontier closing gap or extending lead A B C D E Distance from financial frontier Carbon speed
Co. Gap Carbon speed Convergence Outlook Position
A +708% +311 — DIVERGING Large gap, widening. Financial capacity unconstrained — the constraint is decision-making and strategic allocation.
B +71% −246 3.3 years CONVERGING FAST Behind frontier but closing the gap rapidly. Projected to reach frontier within four years at current rate.
C +46% +31 — STALLED Meaningful gap, drifting away. Capacity to close the gap exists but is not being deployed.
D −3% −0 at frontier FRONTIER HOLDING At the financial frontier, holding position. Maintaining performance commensurate with capacity.
E −21% −10 at frontier OUTPERFORMING Ahead of where financial capacity would suggest, and extending the lead. Genuine outperformance.

Anonymised illustrative cases drawn from Swanstant's 2026 Carbon Frontier model coverage of 103 listed companies. "—" in the convergence column indicates the company is not on a trajectory to reach the frontier at current carbon speed. Real company benchmarks available under bilateral confidentiality.

Across 103 listed companies, we find that not one is financially constrained from further decarbonisation at the margin. Where progress has stalled, the constraint does not appear to be financial capacity; it appears to be decision-making and strategic allocation. Within each industry, the dispersion in climate performance relative to financial position is wide — and widening. We introduce a specific metric, carbon speed, that captures the rate at which a company adds or removes carbon per additional million of revenue. Companies with negative carbon speed are closing the gap to their industry's efficiency frontier; companies with positive carbon speed are falling behind and, on current trajectories, likely to face rising transition risk before it surfaces in their own sustainability disclosures.

The Frontier model is calibrated on 103 listed companies for peer-group comparability; the broader Competitive and Transition Benchmark platform covers 179 companies across 15 sectors and is published at benchmark.swanstant.com. Methodology isolates decarbonisation capacity from observed financial position (profitability, leverage, return on capital) and industry-specific operational structure. Full methodology and carbon speed construction available on request under bilateral confidentiality.

Methodology

Three disciplines, applied consistently to every forecast. And one accountability commitment.

01

Macroeconomic, seasonal, and secular detrending.

Before any climate variable is evaluated, output series are detrended against a panel of macroeconomic and sector-specific drivers — commodity prices, foreign exchange, demand indices, labour cost, energy cost, sector-specific supply chain effects. This prevents the model from attributing to climate what is in fact driven by market conditions.

02

Climate signal isolation.

Across a broad candidate space of climate variables spanning temperature, precipitation, humidity, drought, and — for crops — phenological-phase-specific conditions, we identify the variables that explain residual variability after detrending. The selection process is conservative: only a handful of variables survive validation per regional or sectoral specification. Variable-selection criteria are empirical, based on functional and statistical relevance.

03

Out-of-sample validation.

Every signal is tested on data the model has never seen. Walk-forward validation rolls the training window through time so each year's forecast is generated from data available before that year began. Additional robustness checks include specifications designed to produce null results in the absence of real signal. Calibrated 80% and 90% prediction intervals are evaluated for coverage on held-out periods. Signals that do not pass validation do not enter the published forecast.

Accountability commitment

We update each forecast as new climate observations arrive and preserve previous issues on the public record. Readers can compare any current forecast against its earlier versions and, once the outcome is known, against what actually happened.

The validation results for each published forecast and research finding are disclosed in the relevant sections above. Methodology aligned with the validation standards applied at central bank research departments and institutional quantitative research. Full methodology documentation available on request under bilateral confidentiality.

Coverage

Current coverage, and work in validation.

Sector Geographies Climate variables Status
Cotton US (Georgia, Texas, California) · India (5 clusters, 116 districts) ENSO, dewpoint, heat hours, precipitation Live (seasonal forecast)
Coffee East Africa (expanded panel issuing Q2 2026) · Brazil (in development) Warm-night frequency, ENSO, heat stress In development
Almond California (20-year resilience outlook) · Expanded quality panel (California, Australia, Spain, Morocco) in preparation Bloom frost, water allocation, extreme heat, hull-split conditions, chill hours Live (resilience outlook)
Citrus Turkey (in development) Frost severity, heat stress In development
Natural rubber Southeast Asia (in development) Precipitation, ENSO In development
Cement Global, technology-segmented (2,160 facilities, 4 kiln types) VPD, frost, heat stress, precipitation Live (dose-response)
Steel Global, technology-segmented (958 facilities, 6 production routes) WBGT, VPD, heat stress, frost Live (dose-response)
Pulp & paper Global (in development) Water stress, heat In development
Textiles Asia, Europe (in development) Heat stress, drought (SPEI) In development
Food processing Global panel (15,000+ facilities, panel-validated for margin work) Heat-cooling compound events In development
Competitive & Transition Benchmark 179 listed companies across 15 sectors (fashion, consumer goods, technology, automotive) Carbon speed, financial frontier position Live (annual benchmark)

Beyond our published research programme, our methodology can extend to commissioned work across commodities, sectors, geographies, or decision horizons not currently in the matrix — provided the underlying data supports validation at our publication standard. See engagement for details.

Engagement

How we work with clients.

If our methodology and approach to climate-linked economic impact is of interest, reach out to discuss whether we can help:

contact@swanstant.com · Schedule a conversation

Initial conversations are 30 minutes and designed to establish fit. We typically discuss a specific question on your side — a hedging decision, a supply-chain exposure, a portfolio-level concern, an underwriting question, a research collaboration — and describe how our current coverage or commissioned capability maps to it. Methodology documentation, validation history, and capacity discussion are available to qualified evaluators under appropriate confidentiality.

What we are not

Boundaries of the work.

Swanstant is not a physical climate-risk scoring platform. We do not produce sustainability-reporting outputs. We do not issue ratings used in regulatory disclosure. We do not manage capital, hold positions in the markets we forecast, or provide investment advice. We do not provide adaptation consulting in the form delivered by management consultancies, though our forecasts and methodology can inform adaptation work undertaken by client teams or by their consultants. We forecast specific climate-economic variables at operational horizons and within specific future scenarios, publish the validation, and engage bilaterally with institutional clients who require those forecasts as part of their decision process.

About

Who we are, and how we hold independence.

Portrait of François Souchet
Founder

Swanstant is led by François Souchet, who built the firm to apply quantitative climate-economic research to the operational and financial decisions where climate impact is most material and most consistently underpriced.

Trained as an engineer (classes préparatoires at Lycée Stanislas, civil engineering master's at ESTP, master's in engineering management at the University of Melbourne), François began his career at Accenture in operations and supply chain, advising industrial clients globally on productivity, cost, and operational resilience. He subsequently led Make Fashion Circular at the Ellen MacArthur Foundation (2018–2021), a research institution producing macroeconomic and sector-level analysis on circular systems, where he developed quantitative models of the macroeconomic opportunity of the circular transition and built operational frameworks adopted at scale.

His applied work at Swanstant spans industrial and institutional clients on quantitative climate-economic analysis across commodity markets, heavy industry, and consumer-facing sectors. He is a member of the SMEP Independent Technical Advisory Panel.

The firm's methodology is reviewed by an advisory committee of climate-economic specialists.

Portrait of Ian Banks
Head of research editorial

Ian Banks leads research editorial at Swanstant, ensuring that the firm's published work meets the standard demanded by an institutional readership and that analytical findings are communicated with the rigour their methodology requires.

Ian was previously a researcher at McKinsey covering climate change, contributing analysis across strategy, regulation, and finance engagements. He subsequently led editorial at the Ellen MacArthur Foundation through 2022, overseeing the publication of more than twenty cross-sector research reports on the circular economy. Since leaving the Foundation, he has continued working on its research output as an external editor, alongside editorial collaborations with United Nations bodies and private organisations including institutional investors and asset owners.

Independence

Swanstant is an independent research firm. We hold no positions in the markets we forecast. We do not operate an asset-management business. We do not issue ratings used in regulatory contexts. We accept no compensation tied to the directional content of our forecasts. The only commercial relationship between Swanstant and its clients is through the relationships described above. This structure exists because the decisions our work informs — hedging, underwriting, capex, disclosure — are only as good as the independence of the signal they rest on.