Life insurers routinely model the sensitivities of risk factors to the external environment. Extreme events represent an important new dimension that must also be factored-in. We look at some extreme events and ask what the life sector can do to prepare for the unexpected.
We talk a lot about unexpected and/or un-modeled extreme "events", and rightly so. Failure to build such eventualities into risk models was a contributing factor to the current economic downturn and has led to large losses and company defaults. We look at a selection of extreme events for life insurance, and ask: do we know enough about the dependencies and are we doing enough to build these into our pricing models and risk management frameworks?
From normal to anomalous
There are a number of non-medical sector events that impact not just liabilities in the life industry, but also asset values, premium volume, new business levels and the cost of capital. Economic cycles, for example, influence disability and well-being, seasonal flu epidemics and minor pandemics affect mortality and absence from work, heat waves increase mortality, changes in the capital markets impact asset values, and the financial status of a business affects its ability to continue to write planned business volumes at appropriate price levels. These dependencies are known, monitored and (alongside actuarial projections) are built into life risk analysis, risk management and pricing.
But it is also possible that these events will exceed expectations in terms of severity and scope and/or that other anomalous “events” and interdependencies will arise that are not normally factored-in. We are concerned here with the now frequently termed “unknown unknowns”, as well as with changes in correlation in the more remote areas of the known and “known unknowns”. As regards what is conceivable, the following extreme events and their impact on relevant risk factors must be considered by life insurers and reinsurers:
Such events generally lack historical experience, and even if data does exist, it would be positioned within a very different risk environment to past occurrences. We now look in more detail at some examples of extreme events.
Major economic change
The interdependencies between health and life expectancy and economic standing are complex, but in general, mortality improves with economic development (countered in part by negative social factors such as obesity) and vice versa. A dependency was demonstrated by the convergence in life expectancy in Germany[1] after the reunification of East and West, given no other climatic or genetic population differences. In addition, mortality data showed a marked change in the male:female mortality ratios in Eastern Europe around the economic transition period at a time when the ratios in Western Europe remained approximately constant. A dependency between economic standing and suicide has also been observed in statistics from Japan for example, that show a sharp rise in suicide in 1998 correlating with rising unemployment; both remain high to the present day. As regards the effect of economic downturn on disability, statistics indicate that cases initially reduce amongst the employed, due to job uncertainty, and then increase. Within the self-employed sector, disability cases are more likely to increase at the outset. All these dependencies can be built into risk projections.
An economic change scenario that can act as a starting point for considering the full effect of these dependencies is that of the Swiss Solvency Test (SST)[2]. This SST scenario includes shares, real estate and hedge fund values falling in value by 30% and interest rates curves increasing by 300bp. Aside from the direct effect on liabilities, premium volume falls as life policies are cashed-in, new business falls to 25% of an average year, lapse rates increase to 25%, re/insurance downgrades of three or more notches, and increased cost of capital and reinsurance follow. To give the full potential impact, the resulting effects of such a scenario can be combined with mortality and disability modeling.
Climate change
With its complex impact on the environment and ecosystems, and thus on agriculture and human health, climate change introduces new and evolving risk interdependencies in life (and non-life) re/insurance. As summarized by the WHO[3], the overall health effects of a rapidly changing climate are likely to be “overwhelmingly negative”, for example through increasing frequencies of heat-waves (causing heat stress and increased death rates from heart and respiratory diseases), risk of water-borne disease and coastal flooding, and through longer vector-borne disease transmission seasons and geographical shifts. Climate models are an important tool for monitoring these changes.
Unexpected catastrophic events
The economic and re/insurance industry impact from such events, e.g. the World Trade Center, has been high but not sustained in the long-term. Risk modeling and underwriting guidelines also adjust accordingly. However, these aspects would not prevent other conceivable events from having a long-term impact that could lead to the downgrading of combined life/non-life re/insurers, and/or be followed by a reduction in available capacity and increased capital costs.
Major pandemics
There have been 31 known pandemics in the last 430 years. The potential impact of a major pandemic on mortality is however difficult to assess not just because mortality is disease-specific - what we know today about virulency and the effect on different age groups, for example, is not what we will be faced with tomorrow - but because we are also dealing with an ever-changing risk environment (e.g. due to evolving medical and socio-economic factors, air travel, improved monitoring and control of disease and vaccine programs). As medical resources are strained, a major pandemic would also indirectly increase rates of disability and non-pandemic mortality, and would have wide-ranging social and economic repercussions. Life group policies and guaranteed minimum benefits add significant accumulation risk.
Stochastic modeling can be used to adjust calibrated frequency-severity pandemic models for current risk factors. Consequent shifts in buying behavior and temporary or even permanent economic and social change should also be taken into account.
Are we doing enough?
Re/insurers and regulators can and are taking measures to mitigate the exposure to extreme eventualities that might fall within the sphere of insured risk. As an absolute last resort in the most severe cases, the regulator could reduce guaranteed benefit levels and coordinate the distribution of assets, but this should never need to happen. Effective enterprise risk management is central to avoiding such a scenario. Within a suitable risk management framework, risk models must be based on the most up-to-date data and scientific methodologies to give as accurate a view of risk as possible, and must be continually reviewed. In addition, it is important to look very carefully and critically into the “tail” of loss distributions and at tail dependencies. To that we must add the ability to stand back from the models and to think about new extreme scenarios.
Specifically for life reinsurance at PartnerRe, we use capital models that include a stochastic approach to mortality risk, including pandemic scenarios. We also model extreme scenarios for economic change and incorporate these into our life risk projections. Our non-life modeling expertise adds security through advanced in-house catastrophe modeling and portfolio modeling, and through the use of climate models in respect of non-life risks. All aspects of risk assessment at PartnerRe sit within a robust, integrated risk management framework.