Should we set a limit to diversification?

For asset managers overseeing asset classes with significant pay-off asymmetry, the maximization of diversification has always been a way to minimize the downside risk. It is common practice to minimize the Tail-VAR of the portfolio for carry-oriented strategies like credit or insurance-linked securities (“ILS”). In asset classes with more symmetry, where substantial losses can be compensated by substantial gains, the reasoning is close to opposite. Managers of symmetric pay-off strategies, like equity fund managers are usually looking to maximize the upside.

But is it the right approach to building asymmetric pay-off portfolios? Indeed, the more a single event is diversified, the lower the impact of said event to the portfolio, but the higher the probability to have the event occur. In diversifying a portfolio, the implicit hypothesis is that correlation between negative events should be low.

Correlation has always been a key topic in the credit market that is difficult to assess. Indeed, the correlation between low probability events requires strong hypotheses and complex modelling. The difficulty to predict correlation has been one of the deficiencies leading to the crash of the securitization and financial systems in 2007/2008. In a high correlation environment, diversification should be low while in-depth individual analysis should be emphasized. To the contrary, in a low correlation environment, diversification should be high. Hence, markets with low intrinsic correlation and strong asymmetry should be the ones where diversification is theoretically maximized.

If we consider the Insurance-linked security (ILS) market, it is obvious that it falls in this category. The ILS market’s pay-off is undoubtedly asymmetric with a high probability to receive a recurring premium and a low probability to lose a significant part of your capital. Moreover, correlation between most perils/regions could be assessed as extremely low not to say nil. In the case that an earthquake was to happen in California, it is hard to see any transmission mechanism or common factor which could create a hurricane in Florida.

With such an orthogonality, building diversified portfolios seems to be the right strategy. Knowing that the dominant risk transferred in the ILS market is linked to US hurricane and earthquakes, it seems rational to underexpose the portfolio to those events and to onboard other risks linked to different perils and/or regions.

After few years of high natural catastrophe activity, let’s look at the past performance of the asset class. On average, very few ILS managers performed better than the ILS market itself. Assuming performance in this field could not be linked to any anticipation skill, it is fair to say that the portfolio construction was the differentiating factor.

Why didn’t maximizing diversification work that well?

If we look at a Florida hurricane costing USD 50billion, the expected loss of such an event is usually modelled at 4.5%. Hence, you can easily build a very concentrated portfolio with a 4.5% annual probability to lose 100%. For this risk you are today remunerated at 7.5% over the risk-free rate. Now, to achieve the same return, let’s assume you require diversification to avoid losing 100% in a particular year. You then purchase 50% of the previous risk and need to find an orthogonal risk for the remaining 50%. Let’s take an earthquake in California as an example. To find a 7.5% return you won’t be able to take a risk with a 4.5% probability of occurrence. As per the market pricing today you will need a 5.3% probability of occurrence.

The reason is that diversifying a portfolio is the common ground strategy which creates an offer vs demand mismatch leading to pricing distortion. Hence, by diversifying the portfolio and keeping the same gross return you have to onboard risks with higher probability of occurrence. Consequently, in this example, your probability to lose 50% has been increased while the one to lose 100% has decreased.

Thus, even for portfolios with asymmetry, low to nil intrinsic correlation and no capacity to predict or detect shocks, diversification cannot be the only answer. The cost of diversification must be constantly assessed, and investors should always keep in mind that it could lead to an expensive implicit price and create unwanted distortion in the loss probability curve of the portfolio.

Obviously, another element to consider is the timing of the occurrence of an event. The quicker the event happens, the better the diversification strategy will be (assuming a zero correlation). Indeed, the time will not be enough to enable the concentrated strategy to deliver its stronger carry adjusted for risk. Therefore, the higher the exposure to frequency, the higher the need for diversification.

Moreover, modelling for very low expected probability events starts to be subject to increased uncertainty because there is no frequent and reliable data to perform the study. The primary market has sometimes delivered transactions based on:

• esoteric risks like operational risk or terrorism risk
• difficult-to-model risks like flood or wildfire
• weak data collection like covers for perils affecting emerging countries
• aggregate triggers with a very low franchise deductible opening the door to attrition risks

If we look at the track record of those types of transactions during the past few years, their return is questionable to say the least, with more transactions eroded or exhausted than expected.

This equilibrium between diversification and concentration is more art than science. A qualitative assessment is key even for agnostic strategies like insurance linked securities. Investment processes should not only emphasize diversification maximization but also impose limits to the investment/risk universe. This would help to ensure that lower adjusted returns, uncertainty regarding correlation assumptions and increased modelling risks are not doing more than lightly offsetting the benefit of diversification. In this context, ILS investment strategies that focus on natural catastrophe risks with a level of diversification commensurate with the level of risk should be privileged.

This article is an opinion article and does not constitute investment advice or recommendations. Neither the author nor SCOR Investment Partners assume any liability, direct or indirect, that may result from the use of information contained in this opinion article.