9 June 2025; 13:00-14:00 GMT+1
Presenter: William Shields

Paper to be presented
Title: CreditMetrics™ TECHNICAL DOCUMENT. The focus of Will’s discussion: three related financial formulas that changed the world… and how we could update them for Climate Change
Author: Risk Metrics Group. Internationally designed by Academics and Corporations (including JPMorgan, Bank of America, BZW, Deutsche Morgan Grenfell, KMV Corporation, Swiss Bank Corporation, and UBS), CreditMetrics™ TECHNICAL DOCUMENT was made freely available to the world and underscored a collective commitment to developing an open and evolving standard for good risk management. Could such a thing happen today with Climate Insight? Or is the collaborative spirit that led to the publishing of CreditMetrics™ TECHNICAL DOCUMENT missing in today’s Corporate world where firms see Climate Research / Insight as a key commercial edge, not to be given away for free?
Link to paper: https://ecologyandsociety.org/vol27/iss2/art7/
Three related financial formulas that changed the world… and how we could update them for Climate Change
- Merton 1974:
- PD_i = 1 – N(Z_i) … a firm defaults on its obligations when its assets deteriorate by Z standard deviations of its asset value which also defines its Probability of Default (PD)
- Vasicek 1987:
- Z_i = ρ⋅X + sqrt(1− ρ )⋅ϵ_i …. Z standard deviations can be decomposed into a systematic part X (shared with other firms e.g. deteriorating GDP) and an idiosyncratic part ϵ (e.g. lawsuits)
- cPD_i = N((N-1(PD_i) + Sqrt( ρ ) * N-1(X))/sqrt(1 – ρ )) … rearrange the formula above and you can assess how a firm’s Probability of Default responds to systematic factors X
- When we have enough customers: Portfolio cPD = Sum ( N((N-1(PD_i) + Sqrt( ρ ) * X) / sqrt(1 – ρ )) ) / n … if we have lent to a lot of customers in a portfolio, then we can assume the idiosyncratic risk of each diversifies away and only systematic inputs affect the Portfolio Credit Risk.
- CreditMetrics™ TECHNICAL DOCUMENT 1997: Formalised, expanded and calibrated these equations to standardise Credit Risk assessments for banks.
- Basel Committee on Banking Reform 2004: Following Credit Metrics wide industry acceptance, the above formula was used by international regulators to set the amount of capital many banks must hold against their lending risk: Sum ( N((N-1(PD) + Sqrt( ρ ) * N-1(0.999) ) / sqrt(1 – ρ )) ) * Exposure) … our banks must be able to survive 1in1000 year systemic loss events within their lending portfolios!
Why does this matter? Like E=MC^2 the above formulae changed the world.
The system of formulas above describes a network of related entities across the planet, responding to shared drivers of risk. Climate Change could be considered one of these drivers. So far this framework has not been used as a lever to adjust Banking Lending habits (perhaps partially to blame for the large amount of UK banking lending to fossil fuel projects?), but it has the potential to via the all important rho parameter which describes how correlated a bank’s customers are to each other, as well as how likely firms are to fail in general (the PD_i parameter). Higher PD & Rho means more capital a bank must set aside, which means the less lending of that type they will do. Should international regulators be using this stick more to limit Fossil Fuel lending, particularly given increasing PD & Rho parameters would not stop lending, but ensure that only the most necessary fossil fuel lending was done?
Session Highlights:
Will Shields has shared the document he used and has added notes and comments that arose from the discussion. This document was written by AI with expert review and challenge but was intended more to spark debate, thinking and understanding than to be precise and detailed on what is very large and complex topic area.
Please find the document here: 25.06.09 MCRILG Three Related Financial Formulas That Changed the World ChatGPT.
Overview
William Shields delivered a sharp, systems-focused presentation on how financial models, especially those used by banks to measure credit risk, are shaping and often undermining global climate outcomes. Drawing from his work at Deloitte and leveraging ChatGPT for narrative structuring, Shields argued that the tools we need to drive financial climate reform already exist; we just need to feed them better data and update their assumptions.
The importance of banking risk models, capital, and finance in relation to climate.
“If you care about climate change, you should care about how banks work.”
Banks allocate trillions using risk-weighted models. These models decide what gets built, financed, and sustained — or what gets left behind. And yet it seems most of these models:
- Ignore climate change,
- Misprice carbon-intensive sectors,
- And offer little or no incentives to shift capital to resilient, low-emission alternatives.
Three Core Models that Changed Finance
- Merton (1974) – Defined default as a function of asset decline.
- Vasicek (1987) – Introduced default correlation due to systemic shocks.
- CreditMetrics™ (1997)– Modeled how default risk changes under stress.
- Released openly by JPMorgan,
- Gordy (2000/2003) built on this work to introduce the a key formula which became formed an important risk-sensitive part of Basel II/III capital regulation,
- A rare case of strategic openness yielding a global public good.
Climate and Credit Models: The Disconnect
While financial risk models are elegant and powerful, they often lack climate awareness. Shields explained how climate risk directly affects:
- PD (Probability of Default):
Transition risk (e.g. carbon pricing) and physical risk (e.g. floods) both increase default likelihood. - ρ (Correlation):
Climate shocks hit entire regions and sectors, increasing systemic interdependence.
Yet these shifts are not typically reflected in capital requirements, which means banks still underprice climate risk and overfund carbon-heavy sectors.
Why Stress Testing Isn’t Enough
Shields praised tools like the ECB Climate Stress Test and NGFS scenario sets, but highlighted major limitations:
- Not probability-weighted
- Often yield qualitative responses
- As such often can’t or aren’t used to price loans or allocate capital – though they may be used as blunt instruments to set total sector limits, they are typically not fine-tuned enough to support a gradual shift to an optimal carbon investment strategy
“We run the fire drill. Then we leave the sprinklers off.”
Proposal: A ClimateMetrics™ Model
“We don’t need new regulations. We need new inputs to the formulas we already use.”
ClimateMetrics™ could be:
- Compatible with Basel models,
- Parameterised using NGFS/CMIP6 climate scenarios,
- Transparent, open-source, and sector-calibrated,
- Designed to reward resilience and price systemic climate risk into lending decisions.
The Real Barrier: Institutional Culture
Post-2008 regulation became adversarial:
- Banks now fear sharing models,
- ESG frameworks are monetised and fragmented,
- Regulators require internal control, but not collaboration.
Final Challenge
“We’ve already built the machine. We just need to feed it the future.”
Discussion prompts included:
- Should PD and ρ be climate-adjusted under Pillar 1?
- What about the political shift, particularly in the US, to less risk-sensitive Pillar 1 approaches e.g. Standardised capital methods?
- Who should build and maintain ClimateMetrics™ — banks, regulators, or a third coalition?
- Could capital requirements become a core contribution to a just transition?
Q&A Summary
Q1. Was the release of CreditMetrics™ altruistic or strategic?
Answer: There isn’t a clear answer. According to AI, it wasn’t altruism, but rather strategic openness. JPMorgan released it to shape regulatory architecture and define market leadership. However, the result was a public good, shared globally and still in use today.
Q2. What can we learn from actuaries and insurers about pricing climate risk?
Answer: A lot, especially around Loss Given Default (LGD). Actuaries price physical damage risk routinely. Their tools (hazard curves, exposure maps, structural vulnerability models) could inform LGD assumptions in banking, especially for climate-vulnerable collateral.
Q3. Is there work underway to assign probabilities to climate scenarios?
Answer: Yes, and it’s critically important. Current climate stress tests lack probabilistic grounding. Emerging research (e.g., NGFS, IMF, central bank simulations) is exploring the integration of Bayesian probability weighting; however, this remains a key barrier to climate-informed pricing.
Q4. Regulators don’t trust banks or complex models — how do you sell them on ClimateMetrics™?
Answer: The UK IRB (Internal Ratings-Based) approach is still alive. While regulation is conservative, shifting to standardised models is not a silver bullet. Standardisation often hides risk. The real opportunity is in creating transparent, verifiable extensions to IRB that regulators can co-develop or approve.
Written by Daniel Mardi. Reviewed by William Shields.