Long-Term Value at Risk · Working Paper 05
A portfolio can look diversified yet be concentrated in exposures that move together because they answer to the same underlying driver. The world's largest sovereign wealth fund has measured it, and found its own tools understating the exposure by close to an order of magnitude.
For climate-exposed institutional capital, the residual that standard portfolio construction treats as diversifiable is correlated and systematic. The paper names this under-measured exposure Risk 0. The failure is empirical: covariance estimates driving real allocation are backward-looking, while climate-exposed correlation is structural, forward-looking, and rising.
The empirical case rests on external evidence. The world's largest sovereign wealth fund has arrived at the diagnosis through its own top-down analysis and acted on it, changing its screening instruments. The provider of the foil model has since quantified the missing layer at close to an order of magnitude. Independent whole-economy work triangulates the same range.
The classical tools remain valuable for the risks they were built to measure; the correction is a measurement layer added, not the tools replaced. The paper also proposes a two-register reading of the failure: statistical between holdings, and by analogy between the commercial and human cases for the transition.
Sovereign wealth funds and permanent capital
You steward capital with a multi-generational mandate. You cannot diversify away from the system you are mandated to sustain. The paper documents the failure the world's largest sovereign wealth fund has already measured on its own book: NBIM's top-down analysis returns 19 per cent potential loss on US equity holdings, an order of magnitude above what standard direct-damage models return.
Superannuation and pension funds
You steward pensioner assets whose members cannot exit the economy they retire into. The paper shows why the diversification your covariance matrix books is not the diversification your portfolio holds. Climate-exposed risks co-move under common drivers that backward-looking estimators do not yet capture. Standard diversification does not reach the correlated residual.
Regulators, supervisors, and central banks
You supervise a base of individually-diversified institutions whose exposures are climate-linked. The paper documents why institution-level tools systematically understate a systemic exposure that your macroprudential frame is best placed to see. MSCI's February 2026 macroeconomic Climate VaR finding of up to 9x uplift over direct damage is corroborating industry evidence.
Family offices and long-horizon stewards
You hold multi-generational capital across the same infrastructure, workforces, and geographies over decades. The paper documents why the ordinary tools understate correlated systemic exposure precisely where your long-horizon holdings compound. Where your peers reallocate around exposure that the standard tools price at 2 per cent, the top-down evidence points to 19 per cent.
Risk 0 is the paper's name for the correlated residual that standard portfolio construction treats as diversifiable but is not. For climate-exposed capital, the residual after standard factor and market exposures is systematic, because the same transmission mechanisms arrive across otherwise-unrelated holdings simultaneously. Standard covariance estimators, calibrated on earlier regimes, do not yet see this co-movement. The paper argues that the diversification the covariance matrix books is not the diversification the portfolio holds when climate-exposed risks co-move under a common driver. Risk 0 is the exposure that survives the diversification the portfolio thinks it has.
Standard diversification works only when residual risks after factor and market risk are uncorrelated. The paper argues climate-exposed residuals are correlated by construction because they transmit through shared mechanisms: common physical exposure across regional clusters, common regulatory transmission through extraterritorial regimes, common supply-chain deselection under aligned buyer regimes, common heat-and-productivity exposure across labour markets, common cost-curve exposure under Wright's Law learning dynamics, and common polycrisis reallocation of transition capital. Backward-looking covariance estimators calibrated on earlier regimes do not yet see this co-movement.
By roughly an order of magnitude for institutional investors with permanent capital and system-wide exposure. Norges Bank Investment Management's own top-down macroeconomic analysis returns approximately 19 per cent potential loss on its US equity holdings under a current-policies trajectory. Standard direct-damage Climate Value at Risk models return approximately 2 per cent at the low end. In February 2026, MSCI, the provider whose direct-damage model produced the smaller figure, published a macroeconomic Climate VaR that quantifies the missing layer. Its worst-case scenarios raise losses by up to nine times direct damage.
The paper identifies six transmission mechanisms that make climate-exposed risks correlated rather than idiosyncratic. First, common reallocation of transition capital under polycrisis conditions. Second, common regulatory transmission through extraterritorial regimes. Third, common supply-chain exposure through buyer reselection. Fourth, common heat-and-productivity exposure across labour markets. Fifth, common cost-curve exposure under Wright's Law learning dynamics. Sixth, common physical exposure across regional clusters. Each mechanism transmits the same shock across otherwise-unrelated holdings at once.
The two-register finding is that direct-damage climate models and macroeconomic climate models are measuring different things, and both are needed. Direct-damage models capture asset-level physical and transition damage. Macroeconomic models capture the spillovers through supply chains, demand, and financial system stress. NBIM's own top-down analysis and MSCI's February 2026 macroeconomic Climate VaR both belong to the second register. Independent triangulation from Swiss Re Institute (18 per cent loss of global GDP under 3°C) and the Potsdam Institute (19 per cent reduction in global income under business as usual) sits in the same range.
The paper quantifies systemic value at stake above US$1 trillion within a decade, presented as a floor rather than a forecast and under declared assumptions rather than as a scenario. The estimate rests on the two-register triangulation of macroeconomic Climate VaR against direct-damage models. It is systemic rather than diversifiable because the six transmission mechanisms produce correlated exposure across otherwise-unrelated holdings. For a holder of the whole system, reallocation cannot solve this, because there is no elsewhere to reallocate to.
WP-05 is the portfolio-theory correction the empirical papers require. Where WP-01 measures exposure at entity level and WP-03 dates the cost event that transmits it, WP-05 shows why standard portfolio construction under-measures the correlated, systemic form the exposure takes across a book. WP-06 calibrates the cascade through which readiness transmits at value-chain scale. WP-07 carries the same argument into the return side: where the residual is correlated against exposed holdings, it may run in favour of transition-positioned ones.
Full paper on SSRN: Risk 0: Correlated Systemic Risk and the Failure of Diversification in Climate-Exposed Portfolios
Author page: Joanne Flinn on SSRN
Download this page as a PDF for sharing or board papers.