Climate Risk in Portfolio Construction

Climate risk is a portfolio-construction problem with long horizons, incomplete measurement, nonlinear losses, and substantial model uncertainty. It is not solved by ranking companies on a single emissions field. Physical hazards can damage assets and supply chains; transition policies can reprice carbon-intensive cash flows; liability and technology shocks can affect both. The practical goal is to make exposures explicit, quantify plausible loss channels, and choose a portfolio that remains acceptable across scenarios rather than pretending to forecast one climate path precisely.

Separate physical and transition channels

Physical risk includes acute events such as floods, wildfire, storms, and heat, plus chronic changes in water stress, temperature, and sea level. Transition risk includes carbon pricing, regulation, changing demand, stranded assets, litigation, and technological substitution. Their horizons and mapping methods differ.

Channel	Exposure input	Financial transmission
Flood / wildfire	facility coordinates, hazard maps	asset loss, insurance, interruption
Heat / water stress	operations and suppliers	yield, labor, capex
Carbon policy	emissions, fuel mix, pass-through	margin and demand shock
Technology shift	product mix, capex plans	stranded or growth assets
Litigation	controversy, jurisdiction	provisions and valuation

A company domicile is usually a weak physical-risk proxy. The relevant locations are plants, warehouses, mines, data centers, real estate, and critical suppliers. That creates an entity-resolution and coverage problem similar to alternative-data research.

Scenario analysis is not a return forecast

Scenarios express conditional stresses: if carbon prices rise rapidly, if a regional flood sequence occurs, or if demand shifts, how would cash flows and valuation react? They are useful for comparing portfolio vulnerabilities, not for assigning precise annual expected returns.

scenario loss_i = revenue shock_i + margin shock_i + asset impairment_i
                  + financing-cost shock_i

Translate climate variables into financially meaningful drivers: energy cost, insured versus uninsured damage, production downtime, capex, and discount-rate changes. Maintain several scenarios with different paths and explicitly record assumptions about adaptation, insurance, pass-through, and policy timing.

Construct exposure metrics with uncertainty

Emissions intensity is one metric, but it is not a complete transition-risk estimate. Operational emissions can be reported or estimated; Scope 3 can dominate for some sectors but has much lower reliability. Physical scores depend on geocoded asset coverage and hazard models. Preserve coverage and confidence alongside every score.

Metric	Useful for	Caveat
Financed emissions	climate target tracking	ownership and data quality
Carbon intensity	benchmark comparison	sector and denominator effects
Implied temperature rise	communication	model-dependent assumptions
Hazard-weighted asset value	physical stress	incomplete facility locations
Transition-value-at-risk	scenario comparison	strong model assumptions

Use confidence-weighted exposures or conservative bounds. Replacing missing information with zero makes poorly disclosed issuers appear safe. Replacing it with a sector average may hide tail risks; a prudent system reports both base and adverse imputations.

Add climate risk to portfolio optimization

Climate objectives can enter as constraints, penalties, or scenario-risk limits. For example, a benchmark-aware optimizer can cap the loss under each selected transition scenario while maintaining factor and liquidity controls:

minimize  tracking_error(w) + cost(w) + gamma * climate_uncertainty(w)
subject to scenario_loss_s'w <= loss_budget_s  for each scenario s
           factor, sector, country, and turnover limits

Hard exclusions provide clear mandate compliance but can create concentrated sector bets. Soft penalties provide smoother trading but may miss a target. Robust optimization across several hazard and policy specifications is preferable to tuning weights to one vendor's point estimate.

Avoid disguised factor tilts

Low-carbon portfolios often underweight energy, materials, utilities, and small or emerging companies. Those exposures can dominate returns over short horizons. Neutralize relevant sector, region, style, and commodity exposures when the intended result is a climate tilt rather than a macro bet. Report attribution in parallel: how much performance came from climate selection, sector allocation, factor exposures, and residuals?

Climate signals should complement—not replace—the factor investing framework. A transition-risk score can be an idiosyncratic risk input or a long-horizon expected-return adjustment, but its uncertainty warrants a smaller position in the alpha stack than a mature, frequently observed signal.

Governance and monitoring

Version scenario definitions, data sources, geocoding methods, and model assumptions. Monitor coverage, data restatements, concentration in supposedly low-risk assets, target attainment, and realized climate-event exposures. A portfolio can meet a carbon-intensity target while remaining fragile to flood or supply-chain risk.

Set escalation rules for material controversies or events, but avoid mechanically liquidating on noisy headlines. Risk teams should distinguish a permanent exposure reassessment from short-lived news volatility.

Key takeaways

• Physical and transition climate risks require different inputs and transmission models.
• Scenarios stress portfolios; they do not create precise forecasts of realized returns.
• Location, supply-chain, emissions, and hazard data all have material coverage uncertainty.
• Use robust scenario constraints alongside standard factor, liquidity, and cost controls.
• Report climate outcomes and unintended sector/style exposures separately.