The conventional response to material physical climate exposure in an investment appraisal is to lift the discount rate by a notional figure, typically 100 to 300 basis points, and to call the question addressed. The convention has the virtue of being quick. It has the corresponding defect of answering a question the investment committee was not asking.
This note sets out why we treat the discount-rate uplift as a shortcut that fails when the underlying decision is consequential, and why we have come to prefer modelling physical climate exposure as a yield or output distribution layered on the cash flows. The argument is not against discount-rate adjustments in principle: for screening exercises and portfolio-level scans they are reasonable. The argument is that for committed capital with a multi-year horizon and a real exposure to weather, soil, water or extreme-event risk, the discount-rate uplift hides the decision the committee actually needs to make.
What the conventional adjustment is doing
The mechanics are familiar. The analyst takes a base WACC, adds a "physical climate risk" or "ESG" premium of (for the sake of argument) 200 basis points, and runs the model. Net present value falls. Internal rate of return is unchanged but the headline NPV moves enough to register at the committee. The work, on its face, has reflected the climate exposure.
What the adjustment has actually done is to penalise distant cash flows more than near ones in a uniform exponential pattern. The discount factor at year 15 is reduced more in absolute terms than the factor at year 5. The shape of the penalty is the shape of compound discounting, not the shape of physical climate risk. The two shapes are different.
Physical climate risk does not arrive as a smooth exponential reduction in expected cash flow. It arrives as a distribution: most years are close to long-run mean output; some years are materially below; a small number are catastrophic. The tail matters more than the mean. The discount-rate uplift moves the mean and is silent on the tail.
The question the committee is actually asking
On the engagements where this question has come up most clearly (a vertically integrated agricultural operation with material exposure on one of two operating sites; a closed-environment venture exposed to extreme heat and water-stress events; an infrastructure asset with flood exposure on a coastal site) the committee's underlying question has been a variant of the same thing:
What is the probability that this asset produces a multi-year cash-flow shortfall large enough to breach our covenants, force a recapitalisation, or trigger an unscheduled write-down?
That question is about a distribution, specifically about a lower-tail region of a distribution. The discount-rate uplift does not answer it. A discount-rate uplift cannot answer it: it operates on the expected value, not on the dispersion.
Two committees we have worked with reached the same diagnosis from opposite directions. The first had approved an acquisition on the strength of a base-case appraisal with a 250-basis-point climate premium and was unable, twelve months later, to say how the asset would behave through a one-in-twenty drought year. The second was being asked to approve a development with a 350-basis-point premium and could not see, from the appraisal, whether the premium was protecting against a worse mean outcome or against a tail event that the premium would not in fact cover.
An alternative: model the exposure on the cash flow
The alternative we have come to prefer is structurally simple to describe. The discount rate stays at a defensible commercial number for the asset and its sector. The climate exposure is modelled as a probability distribution on the variable through which it operates: yield, output, downtime, occupancy, or a combined operational index. The distribution is sourced from site-level historical data where available, supplemented by published peril datasets, climate model outputs and operator interviews. The model is run as a probabilistic simulation. The output is not a point estimate of NPV; it is a distribution of NPV outcomes, from which the committee can read the lower-tail percentiles directly.
Sourcing the distribution
The credibility of the approach turns on the inputs to the distribution. We treat this as a three-source exercise.
The first source is the asset's own history. Where the asset has more than five years of operating data, the historical yield, output, occupancy or downtime variance is the starting point. It is the only source that reflects local soil, microclimate, infrastructure and operating practice without modelling assumption. It is also backward-looking and silent on the forward shift in mean and variance that climate change implies.
The second source is the published peril dataset for the location and the hazard. For drought, this means a regional precipitation and evapotranspiration record. For flood, a coastal or fluvial flood return-period set. For extreme heat, a heatwave-day frequency series. These are imperfect (they are not site-specific, and the temporal resolution is rarely what the model needs) but they extend the empirical record beyond the asset's own history.
The third source is a forward adjustment derived from climate-model outputs at the regional scale, applied as a shift in the mean and (where defensible) in the variance of the distribution. We are careful here. Climate-model outputs at the regional scale carry significant uncertainty; we do not use them to generate point estimates but to test whether the historical distribution is likely to underestimate forward exposure, and if so, by how much.
What the simulation produces
The output of the simulation is a distribution of NPVs (or of any other output the committee wants to read). From that distribution, the analyst can pull the metrics the committee is actually trying to use: the probability of a multi-year cash-flow shortfall against a defined threshold; the expected loss given that threshold is breached; the percentile NPV at the fifth and the first centile; the covenant-breach probability under the asset's actual financing structure.
The committee can then size its risk-mitigation choices (ring-fencing, reserve accounts, insurance, structural diversification, exit options) to the tail, not to the mean. This is the substantive shift. The discount-rate uplift framework asks the committee to accept a worse base case in exchange for protection it cannot quantify. The distribution framework asks the committee to take a specific view on tail probability and to size the structural response accordingly.
Where the discount-rate adjustment retains its place
The argument is not that discount-rate uplifts should be removed from the toolkit. Three uses for them remain defensible.
- Portfolio-level scans where the work cannot bear the cost of building a distribution per asset. A uniform uplift is acceptable provided the user is clear it is a screen, not an appraisal.
- Pre-deal screening, where the discount-rate uplift is being used to sort opportunities for further work rather than to support a final decision.
- Comparative appraisal across assets in the same sector and geography, where the uplift is held constant and the analyst is comparing relative attractiveness rather than absolute committed-capital risk.
For committed capital, single-asset, multi-year horizon work where physical climate exposure is material, the discount-rate uplift is a substitute for the analysis the committee is being paid to commission.
Practical considerations
The objection raised most often is that the distribution approach is too data-hungry where site-level history is short or absent. Where that applies, we have used three workarounds: the operator's history at comparable sites as a prior, a regional peer set drawn from published yield or output data with the comparability assumption documented, or a sensitivity exercise across plausible parameters with a range presented in place of a point. In each case the committee has more information than a discount-rate uplift would have produced.
A second concern, raised by internal valuation teams, is that the distribution approach is incompatible with the discount-rate-based framework used for financial statement valuation. This is correct, and it is not a problem. The appraisal supports the investment decision; the valuation supports the financial statements. The two outputs serve different ends, and we do not collapse them.
Closing position
Lifting the discount rate by a notional figure to reflect physical climate risk is a defensible action under specific conditions: screening, scanning, comparative appraisal. As a basis for committing material capital into an asset with real climate exposure over a multi-year horizon, it asks the committee to make a decision it cannot see the shape of. Modelling the exposure as a distribution on the cash flow restores the shape and lets the committee size its response to the actual question, which is the probability and severity of a tail outcome, not the level of the central estimate. The work is more demanding to commission. It is, on the engagements where we have been asked to do it, the work that the committee found it had been intending to commission all along.