The village of Eyam in Derbyshire, central England, was unlucky to discover that the pandemic, then raging 150 miles (226 kilometers) to the south in London, had arrived on its doorstep.
The pandemic was the plague – the year was 1665. The disease had reached Eyam through the delivery of flea-ridden cloth from London to the local tailor, who would then made clothes for the villagers. The fleas carried the plague bacterium and the recipient of the cloth was the first to die.
Within three months another 41 villagers had perished. By spring 1666 a newly appointed rector proposed that, for the sake of other plague-free towns in the Peak District region, the village should self-isolate. A local Earl offered to guarantee food for the town (supplied on a rock at the edge of the village, paid with coins immersed in vinegar – see location below). In June 1666 the villagers reluctantly agreed. Over the summer the plague returned with a vengeance and there were five or six deaths each day. Eventually one third of the population died. But the nearby towns stayed plague free.
Since 2017, in modeling the threat from wildfire on communities in California, the significant new RMS innovation has been in capturing the process of “spotting” (i.e. identifying new outbreaks of fire far from the fire-front). Strong dry winds bring swarms of glowing embers from a raging wildland fire, which can travel long distances. Should these embers settle on shingle roofs, wooden patios or a leaf-filled plastic gutter, a fire will start. Unchecked, the fire will consume a house.
In high-density housing suburbs, wind-driven fire can spread from building to building and consume a whole neighborhood – as happened in the city of Santa Rosa in 2017. And the only way to stop an outbreak is to intervene: to extinguish each ember-ignited fire before it can spread.
Modeling ember ignitions requires sampling the speed
and direction of the wind and also anticipating what proportion of fire-starts get
extinguished before they can spread. Still it only takes one unchecked fire to
burn down the town.
This same process, in modeling “spotting”, is key to anticipating the spread of the new coronavirus into western Europe and North America.
It is an exciting time for the insurance and wider risk management industry. As of January 31, a new asset for the industry – the Risk Data Open Standard (RDOS) has been made publicly available on GitHub, the largest host of source code in the world, with over 100 million repositories and over 37 million users. It is available royalty-free to the public. The RDOS represents a step change in how risk management data is managed and is designed to overcome and improve interoperability issues and ultimately expand business opportunities.
Released as an open source and an
open standard project, this “open” approach adopted for the RDOS represents a
fundamentally new and pioneering initiative for a risk industry accustomed to
proprietary data standards. The industry recognizes the need for openness and
collaboration to tackle data interoperability problems and increase adoption
and innovation, and the “open” approach has been proven to be the best way to
unlock new collaboration among industry participants. The open source method has
worked successfully in many fields, especially when the problem domain is
engineering a standard data exchange (like HTML or JSON) or data processing and
analytics (like Hadoop and Spark).
RDOS is a “superset” of existing standards such as EDM, RDM, OED, and CEDE, that modernizes risk data into an extensible, risk model agnostic, vendor agnostic, single (unitary), and complete containment with all exposures, losses, contracts, structures and domain data in one package.
One of the major drawbacks with current data schemas is they are predominately focused on property risk – existing data models are rigid and difficult to extend to new classes of risk and associated coverage. The RDOS, in contrast, supports the capability to add new lines of business beyond property catastrophe, in a modular way that does not interfere with existing lines of business.
The extensibility of the RDOS schema also empowers open modeling and the ability to customize a personal view of risk. For instance, using the RiskItem entity (see figure below), which already includes the subtype RealProperty, this can be extended to new subtypes that represent new lines of business, like aviation, or offshore energy. Existing attributes, like RealProperty attributes, can be extended to define new attributes that may support advanced analytics or other internal system operations.
With the RDOS, you are not restricted to a particular database
technology, it is a logical data model,
meaning that it provides schema specifications and business rules to
standardize the data model independent of a specific database technology. You
are free to choose the database solution that best meets your organization’s
needs, depending on your access patterns, performance requirements, and scaling
needs. Users can select from several technology options, including Object-Relational,
Relational or a simpler Interchange format such as JSON.
By accessing the RDOS Repository on GitHub, you agree to an Apache Licence 2.0, which is a straightforward, widely used open source licence that grants access, usage and modification rights for private and commercial use.
To contribute to the RDOS, participants simply need to access RDOS via GitHub at this link: https://github.com/RMS-open-standards/RDOS and add their contribution – whether it is creating a new branch or a data subtype, implementing a new feature or fixing feature(s) as desired, and then provide comments so that others can understand your feature. The ongoing development of the RDOS is being guided by the RDOS Steering Committee, which is composed of industry experts from leading companies. The committee meets regularly to review and approve/disapprove RDOS contributions that may be added to the standard.
If you are a developer, a risk modeler, a data analyst – or if you are from outside of the risk industry and want to explore the RDOS, there is a wealth of documentation available and a microsite which explain the RDOS in more detail. If you are already familiar with working on open standards, please explore via GitHub, and come join the party – it is time to create something amazing together.
Natural disasters and other large catastrophes can trigger huge economic losses potentially among multiple insurance lines. RMS, in collaboration with research partner Cambridge Centre for Risk Studies at University of Cambridge, have developed eight template scenarios that model liability clash triggered by natural or man-made catastrophic events.
This research was primarily focused on property–liability clash modeling and was the continuation of the two-year Global Exposure Accumulation and Clash (GEAC) project. GEAC laid the foundations for a risk data schema for the various insurance lines, as well as an approach to modeling insurance clash among life and non-life insurance. Please see here for more information.
The project resulted in liability loss assessments for various scenarios including natural catastrophes, cyber and terrorism (see Figure 1 below). Although these scenarios are extreme, they are well within the realm of possibility and are aimed at stressing the various types of liability insurance.
In the spring of 2003, RMS pioneered quantitative event cancellation risk analysis with a study for FIFA in respect of the 2006 World Cup in Germany. As it happened, SARS was first reported outside China in February 2003, and was rampant throughout the duration of the risk analysis. At that time in London, as in Asia, sensible precautions such as avoiding busy Chinese restaurants was a rational defensive measure. Since the World Cup was scheduled for three years later – the summer of 2006, SARS was not considered as a cancellation risk. Terrorism was the primary risk to which investors in Golden Goal Finance Ltd were exposed.
Thanks to intensive global contact-tracing, and the need for an infected person to be symptomatic before being contagious, the World Health Organization was able to declare the SARS outbreak contained in July 2003. Nearly seventeen years after SARS, a novel coronavirus related to SARS appeared in China over a month ago in December 2019. Whereas SARS had a case fatality rate of about ten percent, the novel coronavirus (2019-nCoV) is more benign. The case fatality rate is currently estimated at just a couple percent. But even this level is highly disruptive, and all risk stakeholders will be anxious over the number of months before 2019-nCoV is contained.
Data – the buzzword of the decade. The world understands its value, but the insurance industry has not only lagged behind in exploiting data, it has also created huge inefficiencies in how it is handled and exchanged. This situation needs urgent attention, but no single company is going to solve this problem – it will take collaboration.
The data that drives risk analytics has proven particularly tricky to handle and leverage. Right now, the standards the industry uses are decades-old property cat schemas – venerable workhorses that took the industry from an almost total lack of exposure data to a relatively high degree of understanding. They transformed how property cat risk is transacted, priced, and managed. But these old formats have run their course and if we want to gain meaningful efficiency, improve profitability, and pursue new opportunities beyond property cat. The industry needs an improved standard.
Fortunately, an intriguing possibility exists. Over the past several years, RMS has researched and built a comprehensive and flexible new data container called a Risk Data Object. This data specification can handle any type of model, any line of business, and any financial terms and conditions.
In 2017, WannaCry infected computers in over 150 countries across the globe, taking out critical functions such as the National Health Service (NHS) in the U.K. One year later, the NotPetya cyberattack brought many household names to a standstill. The pharmaceutical giant, Merck, was reportedly the source of US$1.3 billion of total impact to (re)insurers from the NotPetya attack, 87 percent of which was considered silent exposure. These two major cyberattacks highlighted to insurance carriers the risk of being exposed to silent cyber events and the need to start quantifying and managing that risk.
Regulators have started to
take notice. Since summer 2017, the U.K. Prudential Regulatory Authority (PRA) is
asking insurance firms to provide action plans on how they plan to address
their silent cyber risk. In November 2018, Moody’s announced it will soon start
evaluating organizations on their risk to a major impact from a cyberattack.
Following this, in July 2019, Lloyd’s announced a deadline of January 1, 2020
for all syndicates to start to address their silent cyber risk where “… all
policies provide clarity regarding cyber coverage by either excluding or
providing affirmative coverage.”
NotPetya and WannaCry were just two examples of
costly silent cyber events. As pressure from regulators mounts and cyberattacks
become more common, it is imperative to understand where silent cyber exposure
can be found, and how much it could cost you.
The Risk Data Open Standard (RDOS) Steering Committee that guides and oversees this standard was busy in 2019 working to shape the RDOS, to enable the risk industry to simplify risk management and risk data portability. The RDOS has traveled a long journey and is now proudly an open standard.
Much has been
published already about what the RDOS is. So, for this post I won’t focus on
the mechanics of the RDOS, instead I’ll focus on the “why and how” the RDOS is
open, and how it enables anyone in the industry to use and contribute to the
Just in case you are not familiar with this new standard, I will quickly introduce the RDOS, but the majority of the post will be on diving into the “open” part of the RDOS, and to unpack how we are reusing learnings from existing “open source and open standards” that have successfully created international and industry wide collaborations.
Epidemiologists are disease detectives. The
investigative insights of a forensic epidemiologist are exemplified by Sherlock
Holmes, whose creator, Arthur Conan Doyle, qualified as a medical doctor in
Edinburgh. With limited information, some of which may be dubious and
misleading, epidemiologists search for hidden clues as to the cause of a
disease and its manner of population spread and use statistical modeling techniques
to estimate the degree of disease contagion and the number of cases of
Prof. Neil Ferguson heads the World Health Organization (WHO) Collaborating Center for Infectious Disease Modeling at Imperial College London. His search for scientific understanding using sparse observational data dates back to his theoretical physics PhD at Oxford. Like others trained in theoretical physics, Prof. Ferguson is not shy in making mathematical forecasts that may be at odds with partial data of suspect reliability. Misreporting blighted the Chinese response to the 2002 SARS outbreak.
Mid-January saw the publication of the annual World Economic Forum (WEF) “Global Risks Report” timed to set the agenda during this week’s WEF Annual Meeting in Davos.
With each new edition – and this year’s edition is the fifteenth, inevitably, one first turns to the opening page of the report, to discover the Top Five Global Risks for 2020, in terms of their “likelihood” and “impact”. What has been trending and what has slipped down the chart?