14.1 Questions to Be Tackled from an Actuary’s Perspective

The COVID-19 pandemic and the widespread shutdown of social and economic life in 2020 hit the insurance industry hard. In addition to losses in investments due to falling share prices and lower interest rates in reinvestment, non-life insurers and reinsurers in particular suffered high claims costs especially in travel insurance, business closure and event cancellation. Even if the situation of the industry is not existentially threatening from today’s perspective, the COVID-19 pandemic has impressively shown the negative effects extreme events can have on the financial and solvency situation of companies (see also Frank et al. (2020) and Actuarial Association of Europa (2020)).

To make matters worse, the course of the COVID-19 pandemic is a “natural disaster in slow motion”, with no end in sight even at present. With all the uncertainties of further waves and the effectiveness of the vaccination program the further course of the pandemic and the economic consequences are still difficult to predict.

Management of the crisis for both the private and public sectors is based on the analysis of data and appropriate models and metrics. Therefore, in Sect. 14.2 we first discuss the learnings and their practical application from COVID-19 on models, risk modelling and assessment. We especially share some thoughts on the general insurability of a pandemic in Sect. 14.2.1 including a first assessment of how the risk appetite and the capacity for covering a pandemic might have changed due to COVID-19 and what innovative solutions for a pandemic cover could be expected for the future. In 2020 we especially gained more experience on how to manage an extreme event like a world-wide pandemic and we elaborate in Sect. 14.2.2 on aspects of data quality and information uncertainty, how the massively increased volatility during such a crisis can be handled and, based on this experience, which changes of the general risk management framework Solvency II can be expected for the future. Finally, in Sect. 14.2.3 we summarize some potential future consequences which COVID-19 could have on existing products and how this can be monitored and reflected in their prices and reserves. In Sect. 14.3 we discuss how to improve measures for mitigating risks from a governmental perspective. Again, we first discuss in Sect. 14.3.1 what can be learnt for the management of the crisis, now from the perspective of a government. Finally, we present in Sect. 14.3.2 some thoughts on how the economy can generally be made more resilient for such a crisis, how pandemic risks can be covered and better spread on several risk carriers, what role the capital market and especially the World Bank could have and why we have to especially solve the issue of basis risk to implement an effective risk transfer for a pandemic. In a brief summary in Sect. 14.4 we conclude with the main takeaways and what are relevant learnings after such a pandemic.

14.2 Managing a Pandemic as a (Re)insurer

The COVID-19 pandemic is testing the models, methods and processes (re)insurers have used so far. Let us first consider in general terms how we can deal with new knowledge and uncertainties in assessing risks during an extreme event such as a pandemic. For assessing prices and risks three types of models are typically used: calculation or pricing models, valuation models, and risk models.

The COVID-19 pandemic may have an impact on all three types in different ways. A conscious approach to the respective characteristics and requirements and the questions derived from them for the individual model types are important:

  • Calculation/pricing: Are pricing assumptions still valid? Is there a sufficient margin/return on investment (also in conjunction with a possible adjustment of the allocation of capital costs from the risk model)?

  • Reserving: For which potential losses should provisions already be formed and reported, even under uncertainty? How do assumptions already need to be adjusted, if necessary, also for the expected payments and future projections?

  • Risk: Are the model assumptions still appropriate? Was the exposure fully and correctly considered in the valuation? Are there other dependencies/correlations in extreme events between individual models?

In the following sections we will consider effects and learnings of COVID-19 on all three types of models and the respective processes and measures when applying these models.

14.2.1 Insurability of a Pandemic

The main purpose of (re)insurance is to provide compensation for losses from risks. Insurance is ery effective when covering, e.g., losses stemming from accidents with limited exposures. In this case the main risk is simply volatility and large enough portfolios help to diversify effectively against high aggregate losses. Even local catastrophic events like floods, storms, earthquakes or epidemics are (re)insurable, if on a global level these risks have comparable low frequency, limited size, and, hence, can be diversified against other rare risks in a global portfolio. For such covers, it might happen that a (re)insurer faces rather high losses in a single year but these can be compensated by reserves and equity buffers build up in the previous 5–10 years—some companies even accept longer periods for certain risks and under Solvency II (re)insurance companies have to hold equity to compensate for a 200-years event.

Risk equalisation in the collective or over time no longer works if the risks are systemic or world-wide events have to be considered. A pandemic is a very relevant example for such an event. Obviously, in such a situation the diversification over a world-wide portfolio would not work as effectively as it is necessary for a collective diversification. Assets and liabilities of several lines of business from life, health to non-life are effected. And losses could be so extreme that it is simply not possible to built up sufficient reserves and equity to compensate for the total loss that needs to be (re)insured in such an event (see also Frank et al. (2020) for a more detailed discussion).

Risk appetite and capacity for pandemic risks

With COVID-19 we saw quite impressively that during a pandemic collective risk sharing over different risks within a (re)insurance portfolio does not work. Both sides of the balance sheet, assets and liabilities, were affected, in some cases heavily. And also especially for the liabilities, claims in many lines of business were triggered simultaneously, starting from the obvious life and health businessFootnote 1 to non-life portfolios with main exposures in business closure, business interruption and event cancellation, and other lines of business like, e.g., credit insurance (see also Actuarial Association of Europa (2020) for a more detailed overview and current data can be observed on Roser et al. (2020)).

The problem of the massive potential exposure during a pandemic can be better illustrated by an example from Germany (more detailed data can be found in Destatis (2020), GDV (2019) and GDV (2020)). When only considering a business closure insurance for the hospitality industry sectorFootnote 2 during lockdown we could face a loss of more than €90bn.Footnote 3 In contrast to that in 2018 the annual premium income of the German non-life insurance sector was only €70.7bn, the claims expenditures €52.5bn and the own funds were only €110bn. For such rare events like a pandemic the premium income would increase not more than 2–3% of exposure, i.e., less than €3bn. All in all after such an extreme event the German non-life insurers would be insolvent, if they provided such a cover without a relevant limitation.

In several chapters of this book insurability of a pandemic is discussed coming from different perspectives. Obviously, improving the models does not help, as the problem lies with the simultaneous trigger on different lines of business and both, assets and liabilities, and in aggregate a too large total exposure. Improved data and model quality would only improve the prize of the cover, which is not relevant during the realisation of such rare events. What is important, however, is clarity on the claims trigger itself. Treaty wording has to be strong and clear and the definition of the claims trigger simple but adequate for transparently transferring the risk as promised from the client. In such cases, parametric (re)insurance might be a very valuable too, as discussed in the chapter on “The Legal Challenges of Insuring Against a Pandemic” by R. Hillier. The payout is quicker and supports more effectively the safeguarding of the whole economy. However, the main problem with parametric triggers is always their potential basis risk, i.e. the real claims costs might be materially larger or smaller than the amount paid through the parametric (re)insurance. We discuss this issue in more detail in Sect. 14.3.2.

New solutions and innovations after the pandemic

Greater demand for pandemic cover from clients on the one hand, but lower risk appetite among reinsurers on the other, will drive innovation and new solutions for this cover.

During the crisis (re)insurance companies have developed rather a more limited risk appetite for pandemics. This means that relevant exposure of carve-out or even stand-alone covers for a pandemic will be not available in (re)isurance. To the contrary, the focus will more and more lie on how to enhance the diversification of portfolios, to limit exposures for extreme events, and to better stabilize the whole balance sheet also during an extreme event. This could even mean that several (re)insurers will rather further limit or even try to exclude a pandemic cover for their new products.

On the other hand, the demand for mitigating pandemic risks is now higher than before the crisis for many decision makers in the concerned sectors. Solutions for mitigating or transferring such risks cannot be provided alone by the private (re)insurance sector and solutions will have to include the public sector. This topic is dealt with excellently in the chapter on “Risk Sharing and Stochastic Premia in the Presence of Systematic Risk: The case study of the UK COVID-19 economic losses” by H. Assa and T. J. Boonen. A more detailed discussion from an actuarial point of view on such solutions is given in Sect. 14.3.2.

14.2.2 Risk Management During a Pandemic

The information uncertainty during the pandemic poses particular challenges to management and risk management of a (re)insurance company but also to the management of the crisis under social economic and general economic aspects, and we can learn from the current pandemic how to improve these measures for this crisis and also future extreme events. For (re)insurance it is important to better understand:

  • How can the period of increased uncertainty be managed?

  • How to adjust requirements during the pandemic, if necessary, even with uncertain information,

  • and especially taking into account errors and uncertainties in modelling and assessing the risk situation?

Data quality and information uncertainty

Especially at the beginning of the COVID-19 crisis we had to cope with limited statistical comparability between countries due to different methods used to measure rates of infection and death, as well as different approaches to testing, testing capacity and criteria applied for test eligibility. As a consequence, it was extremely difficult to assess and compare the success of different approaches taken by the countries to contain the pandemic. Also for this reason some research institutes even decided not to look into the data of reported cases at all but only analyse the reported deaths, as they appeared to be more reliable (see for example Imperial College COVID-19 response team (2021)). Even for a single country the data originating from the first till half year are not really comparable to the second wave starting in October for the same reasons, and this still leads to relevant uncertainties in interpreting the observations.

Now, after one year, testing quality and capacity has increased significantly and so the data quality of the reported cases. Still, we have to cope with quality issues during the reporting process and it might be valuable to apply methods as described in “Outlier Detection for Pandemic-related Data Using Compositional Functional Data Analysis” by Ch. Rieser and P. Filzmoser to smoothen the data. However, a major problem remains the attribution of deaths that were directly caused by COVID-19 and not only died with a COVID-19 infection (a more detailed discussion can be found in Ealy et al. (2020)). To solve this ambiguity of potential wrong or missing attributions to deaths caused by COVID-19, it is helpful to analyse the observed excess mortality instead. Reporting and understanding of this data has proven to be of adequate quality for most countries, and hence this is also a robust source for a deeper analysis of pricing and reserving adjustments required due to the pandemic. A valid source for Europe can be found online at EuroMOMO, EuroMOMO (2021).

Managing volatility

The COVID-19 related fluctuations on the capital markets since March 2020 have shown that short-term, temporary volatility can be much higher than long-term volatility. As of mid-August 2020, share prices have already recovered significantly from the previous low in mid-March and are partly back at the level of year-end 2019. The temporary increase in spreads for government and corporate bonds after the lockdown also reduced the market values of interest-bearing securities on the asset side and led to a loss of own funds. The effect of the (temporary) market distortions on own funds was partially mitigated by the instrument of volatility adjustment (VA) under Solvency II, which serves to avoid pro-cyclical behaviour by market participants. The symmetric adjustment factor (“equity dampener”) has mitigated equity stress in the standard formula by up to 10% for the same reasons.

In many (re)insurance companies, the impact and uncertainty of the COVID-19 pandemic have triggered special analyses and sometimes also ad-hoc ORSAFootnote 4 reporting, depending on the requirements as defined in the internal ORSA guidelines and the actual relevance of COVID-19 on the companies’ economic balance sheet. In addition, the supervisors have taken an interest in the current risk situation of the companies and the impact of the COVID-19 pandemic on the risk profile of the companies and have carried out special queries.

Since the development and duration of the pandemic was uncertain from the beginning and is still difficult to predict, scenario analyses in all phases of the pandemic are proven instruments for assessing risks and determining the risk profile. In this process, several scenarios are run, each with a different course of the pandemic, and the effects of the macroeconomic developments associated with these courses on the company’s capital investment, insured portfolio and ultimately equity and solvency capital requirements are estimated.

In this context it is important to provide robust tools to analyse data and transparent models to calibrate and run such scenarios and use the results for deciding what measures are adequate to manage through the current status of the pandemic. Aspects of improving data analysis and models are covered in several chapters of this book, as in “Some Investigations with a Simple Actuarial Model for Infections such as COVID-19” by A. D. Wilkie, “A Mortality Model for Pandemics and Other Contagion Events” by G. Venter or “Epidemic Compartmental Models and Their Insurance Applications” by R. Feng et al. With this input a cascading approach with a crisis intervention team for balance sheet management can provide the relevant guidance for business decisions:

  • Risk drivers assessed as highly relevant are identified with high frequency. For a pandemic, new drivers can be added, such as new business or reported losses of highly exposed segments. And these drivers would then also be able to be taken into account in parallel in the assessment of the risks by running the above mentioned scenarios.

  • Relevant limits are set for these risk drivers and if they are exceeded, measures are initiated by crisis team. The crisis team should also plan regular meetings with high frequency and have been mandated by the Executive Board with the corresponding decision-making powers.

  • The crisis team regularly reports to the board of directors on the development, proposes adjustments to the risk appetite and corresponding measures, and thereby also obtains an adjusted mandate.

In addition to the crisis team for economic and balance sheet risk management issues, it may also make sense to set up another team to deal with purely operational issues such as business continuity, IT and human resources. A representative of risk management, e.g. the RMF,Footnote 5 should also participate in this group and build a bridge between the two groups.

Adjustments to risk models and processes

We have experienced so far that the principle based Solvency II framework enabled (re)insurers quite effectively to manage through this crisis. Currently, the Solvency II Review 2020 is in its final phase and it is a good opportunity to directly reflect learnings of the crisis in adapting the framework accordingly (cf. EIOPA (2020b)). Also in the future, Solvency II will be reviewed with a predefined frequency and new insights and requirements can be reflected. So far, no relevant changes to the standard formula due to COVID-19 are expected. Only topics on extended or improved reporting, e.g. of the ORSA, are under discussion in some of the latest EIOPA consultation papers (cf. EIOPA (2020a)). Again, the chapters on modelling and scenario analysis in this book mentioned above might provide helpful input for how to improve and extend the ORSA report with relevant scenarios for pandemic events.

14.2.3 Reflecting Potential Future Consequences of COVID-19

For the capital market and many non-life insurance covers, most of the effects of COVID-19 were rather immediate and directly observable. In the global economy and in life and health (re)insurance we have to consider long-term effects. Some of these effects are discussed in this book, e.g. , behavioural changes of the population due to COVID-19, others might be relevant to be reflected in pricing and reserving updates of existing products, as, e.g., different demand and consumption with effects on the economy, effects on the health of recovered with longer term impacts and potential triggering of claims at a later point in time (e.g. the chronic fatigue syndrome, effects on the respiratory system, the heart, or neurological illnesses as late effects of COVID-19), or a worsened general health status because of poorer health care during the pandemic. Again, all these effects might only be observed in the future and we will need a close cooperation between data scientists, modelling, actuarial and medical experts. Methods to monitor and model such developments are again discussed in this book in several chapters and from different perspectives, as again “A Mortality Model for Pandemics and Other Contagion Events” by G. Venter and “Outlier Detection for Pandemic-related Data Using Compositional Functional Data Analysis” by Ch. Rieser and P. Filzmoser.

14.3 Managing Pandemics from a Governmental Perspective

For governments and public bodies it might be relevant to learn from the current pandemic

  • how to optimally impose measures like a lockdown, what are effective test strategies and how to implement a vaccination program,

  • how to manage through the sometimes conflicting objectives of minimising both the number of casualties of the pandemic and the economic loss during the pandemic crisis,

  • and finally how to prepare better in advance to mitigate some of the risks by providing obligatory or facultative (re)insurance or risk pools?

14.3.1 Deciding on the Right Measures During a Pandemic

What we discuss on how to implement an adequate risk management during a pandemic in Sect. 14.2.2 is also true for governments. Managing uncertainty is very relevant especially at the beginning of the pandemic and politicians and public servants need to consult closely with experts from all different areas to understand potential options for mitigating risks during a pandemic and to closely manage their effects, both positive and negative. This book features several chapters on strategies how to better and quicker test for infected, optimal strategies to implement a lockdown in the different stages of the pandemic, and already some thoughts on the effectiveness of vaccination, see “Pooled Testing in the COVID-19 Pandemic” by M. Aldridge and D. Ellis and “Diagnostic Tests and Procedures During a Pandemic” by S. Dunbar and Y. Tang, “Changes in Behaviour Induced by COVID-19: Obedience to the Introduced Measures” by N. Badenes-Plà and “COVID-19 and Optimal Lockdown Strategies: The Effect of New and More Virulent Strains” by J. P. Caulkins et al.

After the SARS epidemic in the late 2000s the Robert Koch Institute prepared an analysis of potential future pandemics for the German Bundestag in 2013 (cf. Deutscher Bundestag (2013)). In this work, as in later analyses of the topic, consideration of lockdown as one of the possible solutions to contain a pandemic was completely absent. Also further risk analysis like the Risk Radar of the CRO-Forum in 2019 (cf. CRO-Forum (2019)) still had no reference to a major risk stemming from a potential lockdown during a pandemic. All in all, the research results presented in this book and the open and transparent discussion of potential measures during a pandemic will help the (re)insurance industry and also other sectors like trade and hospitality to better reflect and prepare for measures taken by the government.

14.3.2 Mitigating Economic Risks for Future Pandemics

After COVID-19 we all have to ask ourselves what we can do to become more resilient in future extreme situations to come. Simply applying the same measures as during COVID-19 might not be good enough, as the next crisis might probably be completely different. The next pandemic might affect a completely different age profile (as already did the Spanish flu one hundred years ago), and we might have to manage completely different events like a global black-out, global cyber attacks or, not to forget, the climate change.

Risk cover for future pandemics

As already discussed in Sect. 14.2.1 the private (re)insurance sector will not be able to provide sufficient cover to protect the whole economy against the next pandemic. For such extreme exposures only pool solutions with limitation of the maximal exposure for the entire insurance industry are possible as private-sector insurance solutions for pandemics. Beyond this maximal exposure and also to limit the moral hazard of governments implementing a lockdown without having to compensate for any economic losses, states must provide the major part of the cover. Due to the systemic nature of pandemic risks, the diversification of risks via pools on a global level will only work to a limited extent, unlike natural catastrophe, nuclear power plant and terrorism risks, where already pool solutions exist in Europe, as e.g. the Insurance Compensation Consortium in Spain, EXTREMUS in Germany and an earthquake pool in Switzerland.

A more detailed discussion on this topic, the potential design of such a pool solution using parametric triggers and the role of the World Bank can be found in three chapters of this book: “Risk Sharing and Stochastic Premia in the Presence of Systematic Risk: The case study of the UK COVID-19 economic losses” by H. Assa and T. J. Boonen, “The Legal Challenges of Insuring Against a Pandemic” by R. Hillier and “All-Hands-On-Deck!—How International Organisations Respond to the COVID-19 Pandemic” by M. C. Boado-Penas et al.

The potential role of the capital market

Often the capital market is seen as one possibility to off-load parts of the pandemic exposure of a (re)insurance company. First concepts for such solutions have been tested in the mid- to end-2000s as mortality catastrophe bonds. The experience shows that the appetite in the capital market is rather limited and this solution cannot provide a relevant relief for the total exposure. The main reasons for this limited interest—in contrast to other catastrophe bonds for, e.g., earthquakes or hurricanes—is the expected and, again during COVID-19, observed high correlation of a pandemic loss and losses in the capital market. Only well diversifying alternative investment products have shown a relevant demand from investors.

However, especially for the world’s poorest countries, epidemic or pandemic covers are not affordable at all. To provide a certain economic protection for such countries the World Bank launched in 2016 the Pandemic Emergency Financing Facility (PEF) covering a maximum exposure of US$195.84 m for 64 of the poorest countries. It was actually designed for covering and providing quick support for expenditures during local epidemics like Ebola. During COVID-19 the full amount was paid out (cf. The World Bank (2021)). Until now, the cover has not been renewed by the World Bank, but it might be a very useful tool for providing and effectively protecting the development aid for these countries and hence it might be of political interest to launch a second bond.

Why basis risk sometimes prevents effective risk transfer

The PEF’s parametric criteria are often criticised for being too slow and complicated. The facility, which is based on a set of disbursement triggers, only releases funds once there have already been a certain number of cases, deaths and countries affected by an outbreak. In particular, when infections occur only singularly in a country, the PEF does not provide funds even though there is an obvious need: a classic case of basis risk.

Basis risk has to be considered from two perspectives: a pure economic for effectively hedging against adverse cash flows which should be mitigated, and a regulatory for criteria if and to what extend risk mitigation instruments can be considered under Solvency II:

  1. 1.

    Economically, for a risk transferring instrument it has to be ensured that in most of the cases when a compensating cash flow should be provided by the instrument, this cash flow should also be triggered and paid out in the amount expected. If not then this has to be considered as basis risk.

  2. 2.

    For reflecting basis risk under Solvency II, EIOPA issued a guidelineFootnote 6 to be considered when and how the treatment of risk mitigation techniques in the calculation of the Solvency Capital Requirement with the standard formula needs to be assessed.

As parametric triggers cannot directly use internal portfolio information of the insurer and must approximate the desired effect by publicly available data, it is obvious that there will be always situations where the instrument was originally intended to provide a compensating cash flow but actually does not. In case of a pandemic and an instrument based on parametric triggers covering the mortality risk of a portfolio, this can especially happen, if

  • the age profile of the portfolio and the public data base,

  • the social status and hence the availability and access to medical treatment in the different portfolios,

  • the gender mix, health status or other relevant risk drivers for the mortality rate during the pandemic event.

do not match. As different events might be driven by different risk drivers, it is almost impossible to find a perfect fit for all conceivable situations. And then the parametric trigger should also be transparent, easy and quick to calculate. Effectively achieving both goals is impossible and we have to cope with a certain but hopefully limited residual basis risk when trying to mitigate risks of extreme events with instruments using a parametric design.

14.4 Conclusion: Our Learnings

For us, one of the most important lessons learned was that it is impossible to predict and (re)insure such rare and extreme events as a pandemic. Each new crisis will be different from the last, and we certainly do not have the right risk models and mitigation techniques to safely manoeuvre through this event. For example, in the event of a blackout following a global cyber-attack, it will be the newly established home office environment that will be the Achilles’ heel, rather than the perfect solutions, as was the case with COVID-19. Therefore, we must always remain critical and quickly learn to adapt our measures during a crisis based on new data, robust statistical methods and experts’ insights. Solvency II, as a principles-based risk management framework that builds a more resilient risk culture in companies, helped us stay in control during COVID-19. And the same approach, with appropriate adjustments, will work during other extreme events.

However, we have also learned that the capacity for these extreme events in the (re)insurance market, but also in the financial market, is limited and may not be sufficient to cover all macroeconomic risks. In such extreme events, governments need to step in and stand by the affected companies.

In the event of a crisis, we must act together in a determined and concerted manner. Global disasters cannot be dealt with by individual states or industries acting alone. Especially for the coming climate crisis, we will have to think about how we can tackle the problem across sectors and as a community of states.