Reviewing the fundaments of the volatility adjustment

The risk-correction under the loop

According to the EU Directive 2009/138 and 2019/2177, the volatility adjustment (VA) aims at mitigating the impact of stressed bond spreads on own funds and plays a central role in the Solvency II framework. It is part of the SII 2020 review together with other long-term guarantee measures.

After EIOPA's consultation paper in 2019, a holistic impact assessment is currently taking place until 01 June 2020 to allow EIOPA providing its advice to the European Commission.

Several deficiencies have been identified on the current VA approach. In this paper, we will focus on the misestimation of the risk correction spread and the resulting potential overshooting or undershooting effect. 

Defining the risk-corrected spread

The discussion on the definition of the “risk-corrected spread” is “fundamental” to the determination of the VA, as it addresses its core concept, namely: “correcting for the volatility in bond spreads in a hold-to-maturity portfolio”. Aside, one of the perceived VA deficiencies relates to the exact interpretation of the VA. However, regardless of whether the VA represents a “compensation for exaggerations in bond spread” or “an additional illiquidity premium on assets that replicate the liabilities”, the risk-corrected spread ought to reflect the component of the spread that is not related to expected credit losses. Hence, appropriately defining the “risk-corrected spread” – equivalently, finding the mathematical relation between the bond spread and its inherent credit component – is central to the VA calculation. Thus, this definition would have to be agreed upon prior to embarking on addressing other deficiencies…

Very roughly, one can write:

Bond spread = Expected credit loss component + Liquidity premium,

where we have assumed that all non-credit related risks are reflected in the “Liquidity Premium”. Following the EIOPA terminology, the “Risk Correction” aims to correct for the non-liquidity-related risks in the bond spread. In this case, one would have:

Risk correction = Expected credit loss component

The risk-corrected spread is then defined as:

Risk-corrected spread = Bond spread - Risk correction = Liquidity premium

Finally, the VA is obtained by aggregating various Risk-Corrected spreads across the different investment classes, and multiplying by an application ratio, which aims to capture (i) a correction for the illiquidity of liabilities, and (ii) the duration mismatch between assets and liabilities.

Key desirable features

Before delving into different calculation approaches for the risk-corrected spread (or equivalently the risk correction), one would want to outline the key desirable features:

1. The risk-correction (expected credit loss component) ought to adequately reflect the heightened risk of defaults observed during a crisis (“point-in-time” nature)
2. The framework should incentivise proper risk management
3. The risk-correction should capture the relevant basis risks. In particular, it ought to take into account the investment horizon: risk-corrections for longer-term investments ought to be more “through the cycle” (i.e., free of the effect of the credit cycle) in their nature than shorter-term investments. This is particularly important when there is a significant duration gap between assets and liabilities.
4. The VA ought to maintain its role in mitigating procyclicality and stimulating financial stability (both in base case scenarios and stressed scenarios).
5. Simplicity of calculation is desired

Calculation approaches for the risk-correction

In this article, we assess three different approaches for the calculation of the risk-correction, namely:

• Approach 1: A fixed constant risk-correction
• Approach 2: A fixed relative risk-correction
• Approach 3: An adjusted relative risk-correction

The table below discusses the three approaches and their core assumptions.

Discussion of approaches and mapping against desirable features

The aforementioned approaches all have benefits and disadvantages. In the table below, we compare the desirable features with the considered approaches:

Case study

Figure 1 below illustrates the impact of the three different risk-correction calculation approaches on the VA over the course of the last months (more precisely, we look at the impact on the currency VA in EUR, in accordance with Art. 49 Para. 1 of Regulation 2015/35). The solid line corresponds to Approach 1 (fixed constant risk-correction), and is in line with the VA published by EIOPA in their monthly technical specifications of the risk-free interest rate term structures. Approaches 2 and 3 are represented respectively by the dotted and dashed lines. It is important to point out that in the calculation of this impact assessment, the only difference between the VAs lies in the calculation of the risk-correction. The aggregation approach and application ratios are the same across the three approaches, allowing for a like-for-like comparison. (More precisely, for the dates prior to March 2020, the 2019 EIOPA representative portfolio and corresponding government and corporate bond weights are applied, whereas for the March 2020 dates, the 2020 EIOPA representative portfolio is applied. Note, however, that any changes in government and corporate bond weightings would not materially affect the conclusions.)

Figure 1: Impact of risk-correction calculation approaches on the volatility adjustment

The above figure confirms our prior statements on how Approach 1 provides the highest reactivity to changes in the macro-economic environment, and hence has the most countercyclical impact. Furthermore, we can also see that, in benign economic conditions (prior to March 2020), Approaches 2 and 3 yield similar VAs. However, in a downturn economic environment (COVID-19), Approach 3 allows for an increased countercyclical effect.

Discussion and further developments

In this article, we assessed one of the fundamental building blocks of the volatility adjustment: the risk-correction. We highlighted a number of desirable features, and mapped these against three candidate calculation approaches:

1. The fixed constant risk-correction, as currently used in calculating the VA under Solvency II.
2. A fixed relative risk correction, as proposed in the Consultation Paper on the Opinion on the 2020 review of Solvency II.
3. An adjusted relative risk correction, as assed under EIOPA’s holistic impact assessment of 2020.

The two relative risk correction approaches (Approach 2 and 3) are preferred from a theoretical and risk-management perspective, as they reflect a heightened credit risk during economic downturns. These approaches are hence also more consistent with other accounting standards such as IFRS9 and its US-equivalent CECL, where expected credit losses are accounted for by a “point-in-time” risk metric.

Capturing the heightened credit risk during economic downturns, however, comes at the cost of a reduced countercyclicality. In contrast to Approach 2 and 3, the long-term averages used to calculate the fixed constant risk-correction under Approach 1, imply that any changes in the bond-spread are directly reflected in the risk-corrected spread, which subsequently results in a more countercyclical VA. Approach 3 manages to retain an increased level of countercyclicality to a certain extent, by reducing the relative credit risk proportion of the bond spreads once they go beyond a certain level (e.g., the long-term average spread).

Despite some of the theoretical benefits of Approach 3, it does require the estimation of a number of additional parameters, which could lead to challenges given the scarcity of representative crisis-time liquidity spread data. In practice, the calibration likely relies on a number of strong expert-judgment arguments. In a next phase, the analyses presented in this article could be further enhanced by diving into potential calibrations for the different risk-correction approaches, including:

• Assessments of different time horizons for the LTAS used in approaches 1 and 3.
• Analyses of available liquidity spread data to gauge the size of the downturn coefficients under Approach 3, as well as assessing suitable levels of granularity (e.g. by elaborating on the proposed calibration approaches laid out in EIOPA (2019), pp. 134-146).

Other measures such as the introduction of a “macro-economic VA” (see EIOPA (2019), pp. 150-156) can further tune the counter-cyclical effect of the VA. One might want to further assess the suitability of the three risk-correction approaches in this context.

Going forward, the reassessment of the volatility adjustment framework could be complemented in the risk management system with either (i) an undertaking specific VA (EIOPA (2019), pp. 109-111) or (ii) a “monetary” VA calculation based on a liquidity-adjusted value of an undertaking’s own asset portfolio, as discussed in Meli, de Leval & Garston (2018). Such a self-assessment of the VA ought to be accompanied with appropriate safeguards including: ORSA analyses, credit risk management policy reviews and, regular monitoring and reporting on the current asset allocations (see EIOPA (2019), pp. 102-103).

As a final statement, we would like to re-iterate the importance of liquidity risk management within financial institutions. Whereas the VA aims to incorporate the effect of the long-term investment horizons, it also leads to the fact that asset-side liquidity risk is not reflected under Pillar I in Solvency II. Asset-side liquidity is a crucial component in ensuring solvability, and ought to be a key topic in the risk management of any insurance company, especially in economic downturns. 

References

1. EIOPA (2019), “Consultation Paper on the Opinion on the 2020 review of Solvency II”, 2.4 Volatility Adjustment, https://www.eiopa.europa.eu/content/consultation-paper-opinion-2020-review-solvency-ii_en
2. EIOPA (2020), “Technical Specification of the information request on the 2020 review of Solvency II – Holistic Impact assessment”, 5.1.2. Volatility Adjustment, https://www.eiopa.europa.eu/solvency-ii-review-information-request-national-supervisory-authorities_en
3. Meli R., de Leval D., & Garston G. (2018), “Volatility Adjustment under the loop”, https://www2.deloitte.com/ch/en/pages/financial-services/articles/volatility-adjustment-under-the-loop.html.
4. Loon, V., Frank, P. (2017). ”Empirical studies in corporate credit modelling : liquidity premia”, factor portfolios & model uncertainty.

Contacts

Daphné de Leval, Qualified Actuary (IA|Be & IA), vice-chair of AAE SII Working Group, [email protected], +32.478.230.231

George Garston, Manager Financial Risk & Regulation, [email protected], +41.79.808.85.07

Roger Meli, Head of Risk & Regulation, Azenes, [email protected], +41.41.726.89.19