Serial Clustering Activity around the Baja Peninsula during September 2014

In the past two weeks, two major hurricanes have impacted the Baja Peninsula in Mexico. Hurricane Norbert bypassed a large portion of the west coast of the peninsula from September 5 to 7, and Hurricane Odile made landfall near Cabo San Lucas on September 14th as a Category 3 hurricane on the Saffir-Simpson Wind Scale. A third system, Hurricane Polo, formed Tuesday, September 16 and is forecasted to follow a similar track to Norbert and Odile, making it the third such tropical cyclone to develop in the region since the beginning of the month.

This serial cluster of storms has been driven primarily by steady, favorable conditions for tropical cyclone development and consistent atmospheric patterns present over the Eastern Pacific. A serial cluster is defined as a set of storms that form in the same part of a basin, and subsequently follow one another in an unbroken sequence over a relatively short period of time. To qualify as a cluster, there needs to be measurable consistency between the tracks. This is typically a result of steady, predominant atmospheric steering currents, which play a major role in influencing the speed and direction of tropical cyclones. One example of a serial cluster is the four major hurricanes (Charley, Francis, Ivan, and Jeanne) that impacted Florida during a six-week period in 2004.

During this recent two-week period, the area off the west coast of Mexico has maintained high sea-surface temperatures near 85.1 degree Fahrenheit and limited vertical wind shear, leading to an active tropical development region. A mid-level atmospheric ridge over northern Mexico has provided a consistent steering pattern towards the north-northwest, producing similar observed tracks for Norbert and Odile and forecasted track for Polo. Devastating amounts of rainfall have occurred with these storms. Hurricane Odile dropped nearly 18 inches of rain in areas around Cabo San Lucas, representing nearly 21 months-worth of typical rainfall. This cluster, while generating significant wind and flood damage along the Baja Peninsula, has also caused torrential rainfall in the southwestern U.S., including Arizona, southern Nevada, and southern California. Last week, Phoenix, AZ, one of the hardest hit areas, experienced over 3 inches of rain in a 7 hour span due to the remnants of Hurricane Norbert. This was the most rainfall to occur in a 24-hour period in the city since 1911, an estimated 1-in-200 year event by the National Oceanic and Atmospheric Administration. Significant rainfall and inland flooding is forecast to continue as the remnants of Odile and Polo move inland, which may lead to widespread flood losses and the potential for compound post-event loss amplification.

Using Network Theory to Understand the Interconnectivity of Financial Risk

For today’s regulators, systemic risk remains a major issue. Tracing the connections between financial institutions and understanding how different mechanisms of financial contagion might flow through the system is complex.

Modern finance is a collective of the activities of tens of thousands of individual enterprises, all interacting in a “living” system. Today, nobody truly understands this system. It is organic and market-driven, but the fundamental processes that drive it occasionally collapse in a financial crisis that affects us all.

The increasing risk of financial contagion in the financial industry has triggered a new discipline of research – called “network theory in financial risk management” – which is quickly gathering pace. These valuable studies aim to identify and analyze all possible connections between financial institutions, as well as how their interconnectivity can contribute to crisis propagation.

Later this month, Risk.net will launch the Journal of Network Theory in Finance. This journal will compile the key papers of financial risk studies worldwide to provide industry participants with a balanced view of how network theory in finance can be applied to business.

Papers from the inaugural edition of the new journal will be showcased on September 23 at the Financial Risk & Network Theory conference, which is hosted by the Centre for Risk Studies at the University of Cambridge. I will be presenting a keynote on how catastrophe modeling methodologies can be applied to model financial risk contagion.

Our financial institutions are connected in a multitude of ways. For example, by holding similar portfolios of investments, using common settlement mechanisms, owning shares in each other’s companies, and through inter-bank lending.

As the interconnectivity of the world’s financial institutions and markets deepens, financial risk managers and macro-economic planners need to know the likelihood and severity of potential future downturns, particularly the “tail” events of economic catastrophe. Companies must continually understand how they are exposed to the risk of contagion; many were surprised by how fast contagion spread through the financial system during the 2008 credit crunch.

The regulator’s role in limiting the risk of future financial crises includes identifying Systemically Important Financial Institutions (SIFIs) and understanding what aspects of a SIFI’s business to monitor. Regulators have already pioneered network modelling to identify the core banks and to rank their systemic importance, and can now demand much higher standards of risk management from the SIFIs. Increasingly, similar models are being used by risk practitioners and investment managers.

The studies of network theory in financial risk management, such as those carried out by the Centre of Risk Studies, provide valuable insight for all risk practitioners involved in managing financial risk by providing a robust foundation of science from which to understand, model and, ultimately, manage financial risk effectively.

A Message from Matthew Grant: An RMS(one) and Model Update

As I mentioned in my last blog post, tomorrow RMS is presenting an update on the progress of RMS(one)® and the momentum of our modeling agenda at the investor day for our parent company, DMGT. I want to share some of the highlights of that presentation.

We have committed to our clients that we won’t rush RMS(one) to market and that we are taking the time to get it right. To ensure that RMS(one) meets the needs of all of our clients, we are working closely with our Joint Development Partners (JDP) on an agile program of incremental deliverables throughout 2015. We will demonstrate concrete progress on a quarterly basis to our JDPs and we are confident this flexible, agile approach will subsequently meet the needs of our broader client base and ecosystem partners.

We are taking advantage of new technologies to simplify key elements of the design and improve the performance, functionality, and openness of the system to deliver a highly configurable platform. At a minimum, RMS(one) will be ready to deliver our first HD models, and those of our ecosystem partners, to all of our clients in 2015. We will make it easy for those who wish for continuity to gain access to and use the HD models with RiskLink and or in-house tools.

Client feedback from the beta test program is also being incorporated to ensure that RMS(one) is a platform that supports:

  • A system of record for exposure data
  • All current models and the next generation High Definition models
  • Exposure analytics
  • Loss analytics
  • An ecosystem of risk management solutions

In addition, we continue to expand the RMS(one) Developer Network, which includes client developers, third-party modeling, and application development partners.

Our Modeling Momentum Continues

In Spring 2015 we will release RiskLink 15, which includes updates to our flagship North Atlantic Hurricane Model and incorporates lessons from Superstorm Sandy. Also included in RiskLink 15 is our updated Europe Windstorm Model, which has been enhanced to support Solvency II model validation. Both models are complete and are now in the testing phase.

We have made strong progress on the development of our powerful HD models, which will provide the high resolution needed to model tail risk. In 2015, our new HD models will include:

  • New Pan-European Flood
  • Upgraded Japan Typhoon
  • Upgraded New Zealand Earthquake

In 2016, our HD models will include:

  • North America Earthquake
  • U.S. Flood
  • Earthquake and basin-wide typhoon models for APAC

Our expanding model development team, which grew 25% this year, continues to work closely with clients as we build out our models; for example, we recently announced our technical collaboration with China Re to reduce uncertainty in our China Typhoon and Coastal Flood model.

Our presence in Asia is expanding. We recently opened an office in Singapore and have released new modeling products that include our Economic Exposure Databases and Industrial Clusters Catalogs.

I am encouraged by the continued support for RMS(one) from you as our clients and from our parent company DMGT. I look forward to providing you with more details about RMS(one) in the months to come. We have a full agenda for Exceedance, April 27-30, 2015 in Miami, FL., that includes RMS(one) and our model roadmap and I hope to see you there.

2014 Atlantic Hurricane Season Update: Not Quite 2004

The 2014 Atlantic Hurricane Season is already half over, and with only five named storms in the books and El Niño conditions likely by late fall, all signs are pointing to a below-average season.

Over the last six weeks, organizations like Colorado State University (CSU) and the National Oceanic and Atmospheric Administration (NOAA) updated their seasonal outlooks with similar or slightly reduced numbers, attributing them to a variety of oceanic and atmospheric conditions acting to suppress activity, including cooler than normal sea surface temperatures, higher than normal sea level pressures, and stronger than normal wind shear.

Interestingly, the suppressed activity is not being attributed nearly as much to El Niño conditions as originally thought. Despite high likelihoods that the equatorial Pacific would warm to El Niño levels by late summer, observed El Niño Southern Oscillation (ENSO) conditions were neutral during the July and August period, according to the International Research Institute for Climate and Society.

Such observations have certainly impacted ENSO forecasts for the remainder of 2014 into 2015. As of September 4, the likelihood for El Niño conditions to form during the period from September to November dropped to 55% from a convincing 74% probability back in May. Despite this material reduction, most of the ENSO prediction models still forecast the onset of El Niño by early Fall, peaking during Northern Hemisphere winter 2014-2015 and lasting into the first few months of 2015.

Barring any late season surge in activity, this year will be a far cry from the busier seasons of the past, most notably the 2004 season. Like this year, 2004 was also impacted by weak, neutral El Niño conditions. However, the 2004 season was impacted by a rare type of storm known as Modoki El Niño in which unfavorable hurricane conditions are produced in the Pacific instead of the Atlantic Ocean, resulting in above average activity in the Atlantic.

The most notable U.S. hurricanes during the 2004 season were Hurricanes Charley, Frances, Ivan, and Jeanne. These four events damaged an estimated 2 million properties in Florida – approximately one in five houses – and caused more than $20 billion in insured losses throughout the U.S.

The strongest system to hit land that season was Hurricane Charley. The storm made landfall on the southwest coast of Florida on August 13 as a Category 4 hurricane, causing nearly $15 billion in economic damages – one of the most destructive hurricanes in U.S. history.

Just over three weeks later, Hurricane Frances, a large, slow-moving, but less-intense system made landfall on the east coast of Florida as a Category 2 storm with peak winds of 105 mph.

In early September, Hurricane Ivan developed just south of where Frances formed, intensifying quickly. Moving through warm ocean waters, the storm reached Category 5 strength three separate times before making landfall as a Category 3 hurricane along the Mississippi/Alabama border.

When Hurricane Jeanne made landfall in Stuart, Florida on September 26, it marked the second time in history that one state was impacted by four hurricanes in one season.

At this point 10 years ago, nine named storms had already formed in the basin, with six reaching hurricane status. In total, 2004 saw 15 named storms, nine of which became hurricanes, including 6 that reached major hurricane status (Category 3+).

While this hurricane season shares some common characteristics with the 2004 season, so far, 2014 has been relatively quiet while 2004 was the second costliest Atlantic hurricane season in history.

Managing the Changing Landscape of Terrorism Risk

RMS has released an updated version of its Probabilistic Terrorism Model, which reflects the considerable changes in terrorism risk for Canada, Denmark, Ireland, Italy, and the U.K. as well as the decreased frequency of large-scale-terrorism events for each of the five countries.

To inform the new view of risk, our scientists carried out a comprehensive analysis of global attack and plot data from the past decade. We focused heavily on large-scale attacks – those with the potential to threaten the solvency of an insurer.

The analysis showed that incidents of large-scale attacks have steadily and significantly decreased, which corresponds with a rise in the funding and sophistication of major intelligence agencies in the west.

Our approach to terrorism modeling follows three principles, which have been validated by data on intercepted plots, past successful attacks, and recent intelligence leaks:

  • Effective terrorists seek to achieve optimal results relative to their effort
  • Their actions are highly rational
  • They are highly constrained by pervasive counter-terrorism measures

Of the estimated 200,000 documents taken or leaked by Edward Snowden, one of the most relevant validations of the RMS model is an N.S.A. presentation that explains the routing of international telecommunications traffic. A very significant proportion of international telecommunications traffic is routed through the U.S. and Europe which, coupled with advances in big data analytics and plummeting data storage costs, has made intelligence collection easier and more robust than it has ever been.

 an N.S.A. PRISM presentation explains the routing of international telecommunications traffic

According to available data on the frequency of plots and attacks, the risk of a large-scale attack has been in decline since 2007, but the risk of smaller-scale attacks perpetrated by lone-wolf operatives and homegrown militants remains high.

However, we have learned over the past decade that terrorism risk levels are fluid and can change quickly. With the rise of the Islamic State in Iraq and reports of its successful recruitment of foreigners, as well as ongoing instability in Afghanistan and Pakistan, the risk outlook can change at any moment.

The RMS Probabilistic Terrorism Model incorporates multiple risk outlooks to provide users with the agility to quickly respond to any changes in terrorism risk. RMS is committed to updating its terrorism model as frequently as necessary to provide the most up-to-date, granular, and accurate view of global terrorism risk.

Understanding Aftershock Risk: The 10th U.S. National Conference on Earthquake Engineering

Recent earthquakes in New Zealand and Japan show that aftershock risk can be significant, even though this risk is not explicitly considered in portfolio risk assessment. It is no secret that large-magnitude earthquakes are generally followed by high numbers of smaller magnitude earthquakes and sometimes the ground motions from these aftershocks cause substantial damage to buildings. The science of forward prediction of aftershock hazard is still evolving and assessment of building vulnerability due to mainshock and aftershock sequences is currently an active topic of research.

In order to address this issue with the scientists and engineers, I organized a special session during the U.S. National Conference on Earthquake Engineering with Dr. Nicolas Luco of the USGS and Dr. Matt Gerstenberger of GNS Science. We invited a number of prominent researchers to discuss:

  • aftershock hazard
  • structural fragility/vulnerability before and after the mainshock
  • change in risk due to aftershocks

Aftershock risk is real and consumers feel the pain from increased insurance premiums, as observed following the Tohoku earthquake. According to Insurance Insight, “local earthquake premiums are up 25 percent to 50 percent, say, when compared with normal circumstances.“ So, this issue needs to be addressed to improve our understanding before another large magnitude event strikes.

Professor Fusakichi Omori first observed in 1894 that aftershocks decrease regularly with time. He developed Omori’s law, which is still used for estimating aftershock risk. I presented a paper in a separate conference session on estimation of aftershock risk in Japan following the Tohoku earthquake, which was based on the Omori Law. This is the basis of the RMS® Japan Earthquake Model update. Currently, the USGS is working on developing aftershock hazard based on the Epidemic Type Aftershock Sequences (ETAS) model.

Dr. Ned Field of the USGS, one of the speakers in the session, stated that developing the aftershock model is “one of the strategic-action priorities of the USGS in terms of providing effective situational awareness during hazardous events.”

In the meantime, GNS Science has developed a time-dependent hazard model for continuing the Canterbury earthquake sequence. Dr. Gerstenberger of GNS Science reported that GNS has carried out “broadband ground motion simulations” for a suite of moderate sized aftershocks in order to develop aftershock hazards in the region.

Dr. Luco, the co-convener of the session, proposed probabilistic risk assessment for “post-earthquake mitigation decisions” after the occurrence of mainshock (or any other earthquake).

They discussed three different approaches of aftershock hazard calculations and two approaches for estimating increased collapse probability of buildings due to aftershocks. These approaches can ultimately be synthesized to compute the increased earthquake risk of damage or collapse of buildings following earthquakes.

RMS will continue to work alongside our industry colleagues to improve understanding of aftershock risk.

Assessing the Risk of a Global Ebola Pandemic

With the current outbreak of Ebola in western Africa, as well as the recent MERS coronavirus and H7N9 avian flu outbreaks, the world is becoming increasingly concerned about the risk of emerging infectious diseases and their potential to cause the next pandemic.

As catastrophe modelers, how do we assess the risk of a pandemic?

To understand the potential dangers of Ebola, it’s helpful to look to the framework we use at RMS to model infectious disease pandemics. The RMS® LifeRisks Infectious Disease Model projects the excess mortality risk for existing infectious diseases, like influenza, as well as infectious diseases that are emerging or have recently appeared, like Ebola. When modeling a disease, we first look at two main criteria: the virulence and the transmissibility of the pathogen responsible for causing the disease. We then take into account mitigating criteria, including medical and non-medical interventions.

Virulence

Virulence is a measure of how deadly a disease is, typically measured by the case-fatality rate (CFR), which is the proportion of people who die from the disease to those who do not. The current Ebola CFR is 55 percent. For comparison, the CFR for bubonic plague typically ranges from 25 to 60 percent. CFR for flu is typically less than 0.1 percent.

Virulence

Transmissibility

Transmissibility refers to how likely an infected person is to transmit the disease to another person, and is measured in terms of the basic reproductive number, or R of infection, which is the average number of additional infections one person generates over the course of illness. In order to cause an epidemic, R needs to be greater than 1.

The R for the current Ebola outbreak is greater than 1, and the disease will continue to spread. Past Ebola outbreaks have been estimated to be in the 1.3 to 1.6 range, but have occasionally been greater than 5, which is why there is cause for concern. However, Ebola is less transmissible than many other infectious diseases. For example, measles, which is highly transmissible, has an R of greater than 10 in an unvaccinated environment.

Transmissibility

Societal and Environmental Factors

Societal and environmental factors can play a large role in transmissibility. In this case, societal and environmental factors in West Africa have contributed to the disease’s spread. For example, traditional burial practices in which families wash the deceased can expose additional people to the virus.

However, the risk of Ebola developing into a pandemic that extends beyond the region is low, due to the standard public health and infection control practices in place in many countries globally. Ebola can only be transmitted via direct contact with bodily fluids, especially blood, which means that caregivers are the primary people who might be exposed to the virus. In many countries including the U.S., the general practice is to treat all blood as potential sources of infection, due to experience with HIV and other blood-borne diseases. In quarantine situations, such as those being used with the American Ebola cases in Atlanta, the likelihood of transmission from a single person is miniscule.

Medical and Non-Medical Interventions

Medical and non-medical interventions mitigate the risk of an infectious disease pandemic. Typical medical interventions for infectious disease include pharmaceuticals and vaccines. Often, there is no specific therapy or drug available for new or emerging diseases. In these cases, we model the effect of supportive care, which includes management of blood pressure, oxygen, and fluid levels. As we’ve seen with the current outbreak, supportive care and the access to healthcare can vary substantially, depending on the region or population. With the exception of experimental treatments, there are no pharmaceutical interventions available for Ebola. Experimental Ebola drugs are not applicable to large populations at this time.

If there are high enough immunization rates, vaccines can reduce or stop the spread of diseases like measles or whooping cough. Unfortunately, a vaccine isn’t currently available for Ebola. Ebola outbreaks occur sporadically and are caused by different virus strains, making vaccine development more difficult.

In addition to vaccines and medical interventions, we account for non-medical interventions when modeling the impact of pandemics. Non-medical interventions include quarantines, school closures, and travel restrictions. Various countries in Africa have begun to implement these methods in hopes of stopping the spread of Ebola. However, these types of countermeasures can often be difficult to time or enforce properly. Ebola can have an incubation period from two days to as long as 21 days.

So, what is the pandemic potential of Ebola?

The current outbreak is now the largest outbreak of Ebola to date, and the World Health Organization (WHO) has designated the outbreak as a Public Health Emergency of International Concern. However, while cases will continue to develop, a global pandemic is unlikely. Even if the disease were to spread to other regions of the world, Ebola is still considered a rare disease and the transmissibility is likely to be much lower due to quarantine and infection-control measures, even if the CFR remains high. We have not seen any community transmission outside of Africa, and this is expected to continue. Ebola is a very serious disease, with devastating consequences to impacted communities. As risk managers, we aim to improve understanding of catastrophes such as pandemic disease so that as a society we can be better prepared to mitigate risk and recover from catastrophes.

Rebecca Vessenes contributed to this post. As a Senior Quantitative Modeler at RMS, Rebecca is involved in the development and parameterization of the LifeRisks longevity models. She recently completed the longevity model for Japan and has worked on determining the correlation structure for mortality improvement between countries. Prior to working for RMS, she led the Financial Modeling group at AIR. Rebecca earned a Ph.D. in mathematics from California Institute of Technology and is an actuary with the Society of Actuaries.

A Message from Matthew Grant, Head of Global Client Development

I wanted to take this opportunity to give you an update on RiskLink 15.0, our HD models and RMS(one).

RiskLink 15.0 will be released in early 2015, and will include updates to the European Windstorm and North Atlantic Hurricane models. We are also making good progress on the industry’s first HD models, which include Japan Typhoon, Europe Flood, New Zealand Earthquake and U.S. Earthquake.

We appreciate that you are keen to hear about the progress of RMS(one).

The leadership at RMS and DMGT, our parent company, remain committed to our modeling agenda and to bringing RMS(one) to market. We continue to see support from our clients as we work on finalizing our plans.

The vision for RMS(one) has not changed and we will deliver on our promise to provide a platform that supports:

  • A system of record for exposure data
  • All models, current and new generation
  • Exposure analytics
  • Loss analytics
  • An ecosystem of risk management solutions

While RMS(one) will be delivered in 2015, which is later than expected, the revised timing of the release enables us to strengthen the performance and scalability of RMS(one), improve functionality and thoroughly test RMS(one) before it is released.

There is a DMGT investor update on September 18 and that is when we will share the RMS(one) release plan with you.

Today, I can share that we are making excellent progress with the technology that will power RMS(one) to scale. We are incorporating lessons learned from our beta program, and are engaging with experts, including the technology industry’s leading platform and analytical application specialists, to augment the RMS team.

We are simplifying the user experience to make RMS(one) easier to use. We are creating new applications to expose more of the underlying data and analytical engines, which will enable you to craft bespoke solutions for your business needs.

There is a lot of work going on to ensure that we get this right and we look forward to sharing our next update with you on September 18.

Hawaii Narrowly Escapes Hurricane Landfalls in the Midst of an Active Season

Hawaiians held their breath early this month as not one but two hurricanes made their way toward the islands, following similar tracks. Hurricane Iselle was the second and strongest tropical storm on record to make landfall over Hawaii’s Big Island, and though it caused localized flooding, knocked out power to thousands of homes, and took down trees, it did not cause any major damage or injuries. Iselle was followed closely by Hurricane Julio, the fourth major hurricane to form in the East Pacific Ocean so far this year, which bypassed the islands altogether. The Hawaiian Islands were spared once again.

This was not surprising. Storms such as Iselle and Julio, which tracked directly east to west from the East Pacific, typically become disorganized before reaching the islands due to the cool waters and dry air that lie to the east. Hurricanes approaching from the south represent the biggest threat to the islands, due to the warmer waters and more unstable air to the south.

Prior to Iselle, only two hurricanes had made landfall over the Hawaiian Islands since 1949. In 1959, Hurricane Dot tracked from the south and made landfall as a Category 1 storm. In 1992, Hurricane Iniki formed just inside the East Pacific and tracked west into the Central Pacific. It remained well south of the islands and then curved north, making landfall as a Category 4 storm. Both systems made landfall over the island of Kauai, located to the far west of the Hawaiian Island chain.

Nonetheless, Hawaiians and insurers should keep a watchful eye on the weather in the East Pacific. Hurricane season in the East Pacific basin, which officially runs from May 15 to November 30, has so far been characterized by above average activity with 10 named storms, 5 hurricanes, and 3 major hurricanes. The NHC’s 1971-2009 average by August 7th is 7 named storms, 3 hurricanes, and 1 major hurricane.

Activity in the Central Pacific is closer to normal levels. According to the CPHC, between 1971 and 2013, an average of four or five tropical cyclones were observed in the region each hurricane season, which runs from June 1 to November 30. Activity ranged widely from no cyclones in 1979 to as many as 11 cyclones in 1992 and 1994.

While it is rare for hurricanes to make landfall in Hawaii, events such as Iselle and Julio are a reminder that even in paradise there is potential for natural catastrophe loss.

The California Drought: A Shift in the Medium-Term View of Risk

Indications are growing that there is a shift underway in the risk landscape in California that may last several years, prompted by the ongoing severe drought.

It’s no secret that California is a region prone to drought. History shows repeated drought events, and there is emerging consensus that the current drought has no end in sight. In fact, there are indications that the drought could just be getting started.

The situation could be exacerbated by climate change, which is increasing the rates of water evaporation in western regions of the U.S.

We also learned recently that the groundwater levels in Colorado have been depleted by a “shocking” amount, which affects California as a significant amount of water used in the state’s agricultural industry comes from the Colorado basin.

California’s abundant agricultural industry has been fueled by its high sunshine input and the availability of water from the Colorado basin.The state produces nearly half of U.S.-grown fruits, nuts, and vegetables, according to statistics from the California Department of Food and Agriculture.

The sustainability of the agricultural industry is now in question given the emerging information about the security of the water supply, with long-term implications for food production—and therefore prices. While the threat is not to the California economy as farming accounts for little more than two percent of the state’s $2 trillion economy, implications will be to broader food prices and food security issues, as well as the security of those employed to work in this industry.

From a natural catastrophe perspective, we can expect the severity and frequency of wildfire outbreaks to increase significantly for several years to come if current indications prove true. In addition, we can expect that more areas will be impacted by wildfires.

The insurance industry needs to pay close attention to methods for estimating wildfire risk to ensure the risk landscape is accurately reflected over the coming years, just as it adapted in the late 2000s to a forward-looking, medium-term view of the probability of landfalling hurricanes accounting for multi-decadal cycles of increased and decreased hurricane activity in the Atlantic basin relative to the long-term average – and the subsequent consequences for the medium-term risk landscape.