Harjeet Dhillon is an Executive Director and senior member of the Willis Re Financial and Actuarial modelling analytics practice based in the London office. He is responsible for actuarial reinsurance and capital modelling related analysis across most lines with a particular focus on Trade Credit, Surety and Political Risk for international business.
Felton (Mac) Johnston (FMJ): Insurers can use reinsurance to manage correlations risks. How is this done for PRI/TCI portfolios and transactions?
Harjeet Dhillon (HD): The reinsurance market offers multiple ways to manage portfolio correlations and aggregations. They include proportional covers (Quota Shares), non-proportional (Excess of Loss and Stop Loss) and combinations of the two.
Typically PRI covers will consist of Quota Shares that cede on a pre-agreed basis both premiums and claims with return commissions. Any remaining exposure is then either retained fully or further ceded under an Excess of Loss agreement. Proportional treaties are on an automatic basis—not requiring case by case approval by the reinsurer—thereby increasing the insurer’s ability to accept risk by increasing its capacity with the reinsurer following the fortunes of the insurer.
Excess of Loss treaties can apply on a single risk or a portfolio of risks. They protect the insurer should any of the risks in the portfolio suffer a large loss that exceeds an agreed deductible. Event covers (catastrophe) will protect the insurer in the case of a single event that results in the insurer having to pay large numbers of losses that combine to produce a significant “event loss.” This is particularly useful where losses may have come from a single country. Facultative reinsurance—that is, reinsurance placed on a case by case basis—can also be purchased just for specific transactions that involve large peak exposures. It’s possible to pay adjustment premiums depending on how the limit is utilized over the contract period. Reinsurers will typically limit the total amount of recoveries that can be paid (covered in the period of protection) in the form of “reinstatements.” Each time there is a claim where the limit of cover is effectively eroded, reinsurers may allow the insurer to reinstate the cover up to a certain number of times, which typically involves a payment of a pre-agreed additional premium. In theory, Quota Shares could provide unlimited recoveries unless stipulated otherwise. However they tend to have high loss ratio caps to provide an overall limit on the downside risk to the reinsurer.
Stop Losses, although somewhat rare, provide cover in a similar way to Excess of Loss treaties but relate to total claims exceeding a certain pre-determined level. They can be expressed in terms of loss ratios (losses over premiums): for example, a loss ratio of 25% in excess of 125% would provide recoveries to the insurer in cases where the loss ratio exceeds 125% and up to 150%. It can be a very useful method of risk management particularly in financial crisis/recession type scenarios. These covers tend be highly efficient at protecting the downside, but tend to be the most expensive too.
FMJ: What sectoral concentrations in PRI/TCI portfolios are of particular concern?
HD: A balanced portfolio rewards insurers with stability should specific industry sectors produce unexpectedly poor results. An unbalanced portfolio can create volatility with an overweight concentration in one sector for example. Typically we wouldn’t expect to see portfolios with more than about 20% allocated to the top sector. However, having too much concentration in a small number of sectors isn’t necessarily a bad thing on its own. Underwriting a stable and profitable portfolio is about understanding the key risks for each policy issued. Those concerns should focus on the nature of the underlying transaction and policy limits offered.
One example where forecasts for the next year do not paint an overly rosy picture is the global construction and steel sector. Growth within the construction industry looks broadly flat for the year ahead and steel prices are falling along with slowing demand and overcapacity. However, this is the case for the sector in general. If you pick out some individual economies, you might see a rosier picture.
FMJ: Other than country and sector concentration risks, are there other correlation risks in PRI/TCI portfolios that you identify or that are of concern to insurers?
HD: Country and sector risks have for a long time been at the core of correlation management. When Willis Re undertakes an actuarial analysis on a PRI portfolio, building a picture of the potential loss scenarios, we consider regional correlations where defaults in one country have the potential for contagion within or across regions. For TCI we consider sector correlation to be important in judging whether defaults in one sector will adversely affect others.
Region and sector are high-level categories and are useful when looking at portfolio modelling. However, it’s likely insurers are looking at a much more granular detail when determining correlation effects. Certainly insurers would look at risks for long-term perils (investment related) vs short-term perils (export trade related) in countries too.
While stochastic modelling can help model extreme events, deterministic models offer a great supplement to a robust risk management framework. Stochastic models have at least one variable that can take a range of possible outcomes (e.g., the likelihood of a political risk event in a country). Deterministic models, on the other hand, will use a set of fixed assumptions (e.g., a defined economic scenario leading to pre-determined outcomes). Stochastic models can help explore scenarios that may not have been considered or imagined by relying upon statistical distributions selected to describe extreme risk. Most practitioners will use Realistic Disaster Scenarios to support their understanding of how correlations could flow through a portfolio and , indeed, supervisory bodies and insurance legislation (e.g., Solvency II and Lloyd’s) may require such scenarios as part of a capital adequacy assessment. Sensitivity to growth assumptions in an emerging market economy leading to macro/micro shocks, or the effects of poor policy decisions leading to political instability are prime examples of potential scenarios. These could have ripple effects into other regions and may well be exacerbated by other external events (natural or man-made catastrophes, for example) that are unrelated but put further strain on the government and public perception. Commodity or currency shocks are other scenarios that could be considered.
FMJ: Does the prospect of financial contagion figure in the analysis and treatment of PRI/TCI portfolio risks?
HD: It absolutely does. Financial contagion scenarios are bread and butter in understanding the potential for “catastrophe” losses in a portfolio. Recession scenarios, political risk default and contagion across regions are actively considered.
A high level of financial contagion is likely to lead to more severe crises. A financial crisis in one country can bring about problems in another just as in a set of dominoes falling down. The banking sector is often given as an example where this can occur. Banks that lend across borders are more vulnerable to international shocks and can make markets behave in unpredictable ways.
FMJ: Are commodity price swings considered in your analysis of correlation risks?
HD: Similarly to the above, scenarios like price swings provide the ability to “sense-check” correlation assumptions used in stochastic modelling. So yes, they do figure. Political risk modelling will not match the level of sophistication of natural catastrophe models: not only is it about lack of investment there, but it’s also about the available or measurable data. So we need to be imaginative about the type of scenarios that could happen. The trouble is political risks are difficult to manage. They are unpredictable, arising from so many moving parts in human societies, but the failure to consider them thoroughly can have catastrophic consequences. We have a suite of models and tools available but we know we will not get to a stage where we can model human behavior precisely. We can use these scenarios, though, to help us understand what could happen in the tail.
FMJ: When you look at country concentrations, do you also consider regional concentrations? Can you give us examples of regional risk correlations, or perhaps of correlations among countries whose economies or destinies are closely linked, even if not geographically proximate?
HD: An interesting part of building the framework for a model is finding that balance between complexity and credibility. We may want to consider the most granular level of detail, but for a portfolio view, we consider much broader categories. The mechanics for developing a model that incorporates regional parameters to project a loss and count distribution were presented by Willis employees and published by the International Congress of Actuaries (ICA) in 2006 [Ref: http://www.ica2006.com/266.html]. Through the support of the industry we were able to build a database with records across 180 countries dating back to 1966, which gave us the data to ascribe correlation parameters that described the degree of dependence across regions. It was a significant step in quantifying the potential for claims to aggregate together.
However, you can’t afford to stand still and rely on the past; the world was a very different place 40 years ago and models now need to add a level of prediction to forecast the future. Political risks faced by differing geographical investors (e.g., USA and Japan), add a further level of complexity that can lead to very different claims characteristics.
We’ve certainly seen how political change can spread quickly through a region with the Arab Spring. However, it’s not only geographical proximity that needs to be considered; economies have become much more interlinked and that needs to be considered as well. Emerging markets tend to have closely aligned growth prospects. Macro changes to a single country in the BRICS economies are likely to affect the majority of them with ripple effects across a portfolio. For example, if you consider a slowdown in growth for China, it’s likely this could impact Brazil’s economy or the economies of other emerging markets as investor appetite for these economies is weakened, leading to a sharp reduction in capital inflow.