10 November 2023

Considering Climate Risk Uncertainty in Portfolio Construction

Climate risk has caused significant challenges for life insurance financial risk modelers. There is a lot of noise in the market, with emphasis on either very precise or very complex modelling. However, it is important to also look at the broader, systematic impacts on financial markets exposures; such as yield curves, inflation, credit spreads, risk premia, and asset class returns.

How do you find the balance and a sufficient level of granularity in your modeling? How do you select and assess the impact of different climate models? Are there risks we are missing? Moody's Analytics have created a series of short articles to help those responsible for climate modeling in the actuarial, risk, and strategic asset allocation functions of a life insurance company. These articles begin to address and simplify the complexities of climate change in processes such as the Own Risk and Solvency Assessment (ORSA), stress testing, and strategic asset allocation.

Considering Climate Risk Uncertainty in Portfolio Construction

To date, insurers have used climate scenario analysis for one of two purposes: risk management under the ORSA; and strategic asset allocation, to adjust investment portfolios to account for climate risks. This article presents a simple case study on accounting for climate risks in asset allocation procedures.

Common approaches to derive an optimal asset allocation include Mean-Variance portfolio optimization and the Black-Littermann approach. Both are built upon future return and risk assumptions. Physical and transition climate risks impact economic output, and are therefore assumed to shift the expected returns across asset classes, regions, and sectors. Mean-Variance and Black-Littermann aim to find an optimal trade-off between the expected returns and expected risks of an investment portfolio. Therefore, changes in expected returns and risk estimates translate into changes in the asset class, regional, and sector allocations.

The following example focuses on the sector return deviations within European equities, and considers three of the NGFS scenarios: Current Policies, Net Zero 2050, and Divergent Net Zero. Figure 1 indicates that the expected return deviations over the next ten years are relatively small under the Current Policies scenario. However the deviations widen significantly when looking at the Divergent Net Zero and the Net Zero 2050 scenarios. In these scenarios, some sectors will yield more than two percentage points less return per year over the coming 10 years, compared to the baseline return assumptions.

Figure 1: Annual expected return divergence for European equity sectors, for three NGFS climate scenarios, compared to a climate unaware baseline. The analyses cover a timeframe of 10 years.

The Black-Litterman model allows us to account for uncertainties around the return projections for the three climate scenarios. We choose this approach to demonstrate the impact of climate scenario conditional return expectations on the sector allocation in a European equity portfolio. Figure 2 depicts the sector deviation from the market cap weights for a standard calibration of the Black-Litterman framework.

Figure 2: Deviations of sector allocation for European equities from the Black-Littermann model for the NGFS climate scenarios - Current Policies, Net Zero 2050, Divergent Net Zero. Market cap weighted benchmark allocation: MSCI Europe. The sum of all deviations is zero for each scenario.

Looking back to Figure 1, the expected sector returns in both the baseline scenario and in the Current Policies scenario are fairly similar, resulting in optimal sector weights that hardly deviate from the market cap weights. In contrast, the Net Zero 2050 and the Divergent Net Zero scenarios assume a significant increase in carbon taxation, resulting in a pronounced decrease of the expected returns for carbon intensive industries compared to others. Most important, in both scenarios the range of expected return reductions is vast, indicating that some sectors suffer a lot more than others. In both scenarios, the worst performing sector is expected to yield approximately 1.5 percentage points less than the best performing sector over the next 10 years.

Using the Net Zero 2050 and the Divergent Net Zero scenario return assumptions in the Black-Litterman model, results in a strong sector shift away from carbon intensive sectors. While the magnitude of the sector deviation is influenced by the parameterization of the Black-Littermann model, the relative sector shifts are unbiased. The wide range of expected return deviations across sectors is the key driver for the significant change in sector weights under the Net Zero 2050 and the Divergent Net Zero scenarios.

Conclusion

The results from this simple study show that going forward, asset allocation procedures need to account for climate risks to deliver optimal risk/return tradeoffs. While this study only focuses on the sector allocation within equities, it is crucial to account for climate risks on the asset class, regional, and sector level–a serious challenge for investment professionals.

Climate adjusted stochastic simulations and scenario analyses can help identify risks in existing portfolio allocations and build climate resilient investment strategies.

For more insights on this topic, and to listen to the climate risk for insurers podcast series, visit https://www.moodysanalytics.com/microsites/climate-risk-for-insurers

Sponsored by
Latest Stories
  • Global temperatures crossed 1.5°C climate benchmark in H1

    17 July 2024

    Gallagher Re report explains why global heat extremes are worrying

  • Regulatory action lags behind nature crisis threatening economy, says WWF

    17 July 2024

    The NGO publishes its deforestation guide for regulators, central banks and supervisors

  • Cytora teams up with Gamma Risk to map out commercial property risk for insurers

    17 July 2024
  • UK car insurance premiums continue to drop

    17 July 2024
  • MS&AD launches AI home analysis with help from UK AI developer Tractable

    17 July 2024