Actuarial news and views from Cape Town and beyond

Leave a comment

An Actuarial Look at Climate Change

I found this article from 1999 by an actuary commenting on the involvement of actuaries and insurers in the debate on climate change.

It was interesting for me to see how he felt that actuaries had a very big role to play in the studies on climate change. At that time, some of the major insurers in the US had already been part of the major climate change discussions that took place in Kyoto and Geneva as well as being part of the Intergovernmental Panel on Climate Change (IPCC).

This he however felt was not enough. At that time, he saw the value of insurers and specifically actuaries getting more involved in research on climate change. With climate conditions changing, this has major effects on the probabilities of adverse events occurring which actuaries would have to build into their models for the future.

There had been much doubt on the science on climate change where he outlined how actuaries collaborated with climatologists to create better models of global weather patterns and what factors affect them. Climatologists have the knowledge on the physical factors affecting the weather isolation but what he adds is that actuaries have the analytic skills and expertise of modelling events that have a multitude of factors affecting them.

The article also discusses some of the effects and studies done on the effect on mortality as a result of climate change. All in all, it is quite refreshing to see how actuarial work can (and does) help save the world. peara



The ASSA2008 Model

The ASSA2008 model was released in March of 2011, and is the most recent version of the ASSA AIDS and Demographic model to be released. The most significant changes to the previous version of the model are greater estimates of the impact of antiretroviral treatment in recent years, as well as significant increases in levels in condom usage over the last decade.

There was a link to the model on the Actuarial Society’s website, however it has been taken down. There are instructions on how to obtain the model here:

I thought it would be interesting to point out that there are a number of shortcomings in the ASSA2008 model. These include:

1. many of the empirical assumptions about sexual behaviour have limited empirical basis
2. HIV transmission probabilities are calculated on an annual basis, making it impractical to allow for transmission dynamics that operate over short periods due to the high level of infectiousness during the first few months of HIV infection.
3. problems with calibration to some of the age-specific calibration targets, for example the model treats the life expectancy quite conservatively
4. the model does not allow for the effect of recent changes in guidelines for ART initiation in adults and children, and the effect these have on MTCT (mother to child transmission) rates
5. the model does not allow for new PMTCT (prevention of mother to child transmission) guidelines, particularly provision of extended NVP (nevirapine) prophylaxis to children breastfed by HIV+ mothers
6. the model does not allow for the effect of the campaign to promote male circumcision
7. the model does not allow for the effect of recent HIV testing campaign (12 million tested in 2010/11)
8. the model does not allow for new strategies that may be introduced in the future (microbicides, PrEP, earlier ART initiation)
A new ASSA model is being developed using the THEMBISA model and will address many of the shortcomings of the ASSA2008 model.

1 Comment

SKA – the Ultimate Big Data Challenge

Big Data is one of the latest buzzwords being used in the field of statistics. Due to the sheer volume of data being produced by current data collection methods, new statistical techniques are needing to be developed in order to handle it. People’s willingness to share information constantly through the use of mobile devices allows agencies to receive constant streams of data allowing them to track anything from sales patterns to the location of potholes.

As stated in the first of the referenced articles, actuaries are ideally placed in order to tackle these challenges, as they have deep statistical knowledge, added to the fact that predictive models are their bread and butter. The use of predictive models is expanding away from predominantly financial spheres, and into the spaces of medicine, law and sport.

The latest Big Data challenge is that posed by the Square Kilometre Array (SKA). The SKA is based in southern Africa, and it is an attempt to build the world’s largest and most sensitive radio telescope. According to IBM, “the aperture arrays and dishes of the SKA will produce 10 times the global internet traffic, but the power to process all of this data as it is collected far exceeds the capabilities of the current state-of-the-art technology.”

In order to deal with this amount of data, new computer systems and analysis techniques need to be developed. The benefit of this is that their use will overflow into other systems and will effectively leave a “legacy” for future analysis of Big Data.


Models of infectious diseases

This post is intended to provide some interesting information about one of the less common types of model we’re likely to find in actuarial work. I will attempt to describe why infectious disease models are relevant to the reader and highlight some innovations in their design and usage.

What is epidemiology and why is it relevant to actuaries?

“Epidemiology is the study of the distribution and determinants of disease in human populations” writes Mark Woodward in his book on the subject. “The essential aim of epidemiology is to inform health professionals and the public at large in order for improvements in general health status to be made.” [3]

Medical scheme administrators, as purchasers of healthcare, are interested in the cost-effectiveness of treatments and their efficacy on individual patients. They are also interested in the relation between individual treatments and population-level disease dynamics because this affects their future claims burden. Lastly they are interested in public health policy to the extent that it affects regulation and influences long-term disease trends. Life offices also have an interest in significant trends in the burden of disease, especially HIV in South Africa.


Continue reading

Leave a comment

Challenges in modelling cyber risk

A challenge currently facing insurers is the difficulty in modelling the risk of cyber catastrophes. I found one article that quotes an insurer as saying that cyber risks are just too systematic and large to insure, and that government intervention is required to make it feasible. Cyber catastrophe risks are different from conventional catastrophes such as natural disasters because the effects of cyber attacks can be global.

I wasn’t entirely satisfied – after all, there are insurers currently offering cyber insurance – so I searched further to find out more about the challenges involved in cyber insurance and possible ways to overcome them. This article explains in greater detail some of the difficulties in modelling cyber risk, and also outlines some techniques that might help insurers better understand cyber risk.

This second article agrees that the main problem with cyber risk is that individual policies cannot be considered to be independent, since cyber attacks could potentially affect a wide portfolio of policies and result in enormous losses. It explains why this is so: IT systems are highly interconnected, so that attacks on one part can affect many others; and it also describes the ‘geography’ of cyber risk to arise from the use of common platforms that share the same vulnerabilities, so that many businesses across industries could all be affected if a single platform’s vulnerability is discovered.

A further problem is that, because cyber threats have only existed for a few decades, there is not a large amount of data available – particularly for extreme events. This problem could be compounded by companies being unwilling to disclose security breaches to their IT systems, so the available data may be incomplete. A further problem is that it can be difficult to determine the actual loss that arose from events – it is difficult to quantify in monetary terms, for example, exactly how much damage a virus causes. I think these issues with data may not be specific to cyber insurance, and we may have to confront similar issues with data one day should we find ourselves designing innovative products.

Large, correlated events are not new to insurers, and catastrophe modelling techniques have been developed for other incidents, such as extreme weather events.  The authors describe the framework of catastrophe models as follows:

“The framework of the models is made up of a large taxonomy of both historical and simulated scenarios of varying magnitudes and frequencies, a hazard model that provides the footprint of each scenario, and the vulnerability of the assets at risk, which together generates an estimate of the potential financial loss.”

It could be possible to model cyber catastrophe risk using the same framework.

The existence of IT systems that service many businesses is also not an entirely new problem; the situation is analogous to major banks that service a huge number of businesses. If these banks fail, the entire economy would be affected, so there has been work to better understand the interdependent risks resulting from bank failure. One technique that could be borrowed from this work to model cyber risk would be to map out the relationships between producers of these important IT systems and users of the systems in a network model, in order to better understand which risks are correlated.

The authors conclude by proposing that if businesses themselves diversified cyber risks – for example, by avoiding industry standard software – and if insurers could better identify interconnected risks and place limits on losses arising from them, insurers would have the capacity to provide the protection that businesses are seeking.

I think this view is more reasonable than the claim that cyber risks cannot possibly be profitably be insured, although it does acknowledge that there is still work to be done.


Attitudes towards model risk

I found this article, which discusses the different ways in which people view and use models, given the limitations that most models have.

Three general limitations of models are identified in the article: models are only approximations to complex problems and as such may miss some relevant features; sensitivity of results to assumptions cannot in general be validated; and there is a strong reliance on past data, despite the concern being with the future.

People have different attitudes towards using models in light of their limitations. By considering two aspects of attitudes towards models – confidence that we can effectively overcome uncertainties in order to make good decisions, and belief that models have a place in the decision making process – the authors have created a two-by-two matrix, identifying four types of people:
  • Intuitive Decision Makers, who believe that gut instinct and market knowledge are superior to model usage
  • Confident Model Users, who believe that models should be used to optimise decisions
  • Conscientious Modellers, who are very technically accurate, and because of this, uncomfortable with the way uncertain results are used in the decision-making process
  • Uncertainty Avoiders, who believe we are not able to make good decisions under uncertainty

Different attitudes towards using models to overcome uncertainty

As actuarial students, I think we are very much exposed to the Conscientious Modellers spectrum of this matrix. I think it’s possible, especially if we are usually surrounded by like-minded people, to forget that not all stakeholders in the business world will view risk and our ability to handle it in the same way that we do. We may well be confronted by people who fit the stereotypes depicted by each of these quadrants at some point in our careers.

The authors’ view is that all four of the perspectives are needed:

“Conscientious Modellers, possibly to their chagrin, need the operational focus of Confident Model Users (to attract investment in the model), the scenarios imagined by Uncertainty Avoiders (to challenge long-held wisdoms) and the survival instincts of Intuitive Decision Makers (to ensure model use does not lead to commercial disadvantage). At the same time, Conscientious Modellers can use the model to challenge Intuitive Decision Makers, by demonstrating “what you have to believe” for the model to be consistent with intuition. Such a challenge reveals management’s implicit assumptions and enhances accountability.”

I think that none of these four views are optimal individually, and in reality, one needs to be aware of the core ideas behind each of the four perspectives in order to responsibly and effectively use models. However, I think that, especially as actuaries, we should lean towards the right of the matrix (after all, if we don’t believe models have any predictive value, why did we study a course dedicated to them?). I think that ‘Uncertainty Avoiders’ take things too far, and that while it may be good to employ some degree of intuitive decision making, it should still be guided by data and models. I think a balance needs to be struck between the technical awareness of the Conscientious Modellers, who are possibly not pragmatic enough to make tough decisions under uncertainty, and Confident Model Users, who might place too much faith in the results of models.

The authors conclude by saying that communication of model uncertainty needs to be more transparent. They say that currently, admitting to high uncertainty would be seen as unacceptable. However, risks should not be borne only by the technical experts who build and use models, but rather should be openly communicated and shared by other parties in the decision making process.