Cambridge Centre for Risk Studies Cambridge Global Risk Framework

CAMBRIDGE GLOBAL RISK OUTLOOK 2018 THREAT PROFILES Cambridge Centre for Risk Studies acknowledges the generous support provided for this research by the following organisations:

The views contained in this report are entirely those of the research team of the Cambridge Centre for Risk Studies and do not imply any endorsement of these views by the organisations supporting the research, or our consultants and collaborators. This report is not intended to provide a sufficient basis on which to make an investment decision. The results of the Cambridge Centre for Risk Studies research presented in this report are for information purposes only. Any commercial use will recquire a license agreement with the Cambridge Centre for Risk Studies.

Paper Citation: Cambridge Centre for Risk Studies, Cambridge Global Risk Outlook 2018 Threat Profiles, February 2018.

Copyright © 2018 by Cambridge Centre for Risk Studies.

Cambridge Centre for Risk Studies University of Cambridge Judge Business School Trumpington Street Cambridge, CB2 1AG United Kingdom [email protected] www.jbs.cam.ac.uk/risk Cambridge Global Risk Outlook 2018 Threat Profiles

Part A Finance, Economics and Trade Risks

1 Market Crash ...... 3 2 Sovereign Crisis ...... 8 3 Commodity Price Shock ...... 11 Part B Geopolitics and Security

4 Interstate Conflict ...... 16 5 Terrorism ...... 23 6 Social Unrest ...... 26 7 Civil Conflict ...... 29 Part C Natural Catastrophe and Climate

8 Earthquake ...... 33 9 Tropical Windstorm and Temperate Windstorm ...... 37 10 Tsunami ...... 40 11 Flood ...... 42 12 Volcano ...... 45 13 Drought ...... 49 14 Freeze ...... 52 15 Heat wave ...... 55 Part D Technology and Space

16 Nuclear Accident ...... 57 17 Power Outage ...... 61 18 Cyber Attack ...... 63 19 Solar Storm ...... 66 Part E Health and Humanity

20 Human Pandemic ...... 70 21 Plant Epidemic ...... 75

Cambridge Global Risk Outlook 2018 Threat Profiles

Cambridge Global Risk Outlook 2018 Threat Profiles This document outlines the methodology of the threat models used to generate the Cambridge Global Risk Index for each representative city in the global economy. It provides for each threat:

(a) Threat Description - an overview of the threat (b) Mapping the threat – data sources used to create a world map of the geographical variation of the threat, and definitions of the Threat Assessment Grade (TAG) used to categorize the threat to each city. (c) Local Impact Severity Definitions – descriptions of the representative scenarios that affect individual cities called Local Impact Severities (LIS) (d) Quantifying the Threat – estimates of the likelihood of cities experiencing the defined scenarios (e) Vulnerability Assessment – characterizes the vulnerability of city economies to the threats, and describes how cities are differentiated in the analysis (f) Consequence Analysis – describes the methodology and sources used to derive the economic impact to cities following potential events

An overview summary of main data sources is provided below. The threat assessment process uses these main data sources to derive frequency, severity and likelihood for each threat and city. Alternative sources as described in this document are used for additional context and validation.

Threat Main Hazard Source Data Cause Uncertainty Part A Finance, Economics and Trade Risks Market Crash IMF Banking Core-Periphery Designation; MSCI Index Volatility Man-Made High Sovereign Crisis S&P Sovereign Credit Rating Man-Made Medium Commodity Price Shock World Bank commodity price history Man-Made Medium Part B Geopolitics and Security Interstate Conflict Global Firepower Index; Uppsala Conflict Data Program Man-Made Medium Terrorism IEP Global Terrorism Index Man-Made Medium Social Unrest Economist Intelligence Unit Social Unrest Index Man-Made Medium Civil Conflict JRC EU Commission Global Conflict Risk Index Man-Made Medium Part C Natural Catastrophe and Climate Earthquake NOAA Significant Earthquake Database Natural Low Tropical Windstorm and UNEP/DEWA/GRID-Europe Global Risk Data Platform Tropical Natural Low Temperate Windstorm Cyclone Data SSCC Tsunami Laboratory; NOAA JetStream Tsunami Locations & Tsunami Natural Medium Occurrences Flood UNEP/DEWA/GRID-Europe Global Risk Data Platform Flood Data Natural Low Volcano Large Magnitude Explosive Volcanic Eruptions Database Natural Medium Drought UNEP/DEWA/GRID-Europe Global Risk Data Platform Drought Data Natural Medium Freeze Extreme Temperate Events Research; EM-DAT Natural Medium Heatwave Extreme Temperate Events Research; EM-DAT Natural Medium Part D Technology and Space Nuclear Accident World Nuclear Association Reactor Database Man-Made Low Power Outage Nation Master Electrical Outage Days; World Bank Electrical Outages Man-Made Medium Cyber Attack Cambridge Centre for Risk Studies/RMS Cyber Risk Landscape Man-Made Medium Solar Storm IGRF/DGRF GeoMagnetic Coordinates Natural Medium Part E Health and Humanity Human Pandemic Institute of Zoology: Global Trends in Emerging Infectious Diseases Natural Medium Plant Epidemic CABI Plantwise Natural Medium

Cambridge Centre for Risk Studies 2 Cambridge Global Risk Outlook 2018 Threat Profiles

Part A Finance, Economics and Trade Risks

1 Market Crash

1.1 Threat Description Financial crises have historically been one of the main causes of destruction of economic output. Financial crises have occurred throughout history, and inevitably cause credit shortages which deny businesses and markets their life blood to operate. Credit crises lead to downturns in economic output, can cause many businesses to fail, and occasionally result in long term recessions. The market crash model makes use of the research carried out at Cambridge Centre for Risk Studies into financial catastrophes, and the publications, models, and datasets from that research programme.1

We define a market crash as a peak-to-trough drop in the primary index of that country’s leading stock market by more than 10% in less than 6 months. Shocks of less than 10% are generally termed as a ‘market correction’. If markets have reduced growth or stay stagnant for multiple quarters it is known as a bear market. Traders sometimes refer to a crash of over 20% as a ‘black bear’, and a crash of over 40% as a ‘brown bear’. When a country’s economy has negative growth for two quarters it is termed a recession. Most financial crises are followed by a recession for some period of time, with the most severe crises causing recessions that can last a decade or more.

Market crashes have occurred throughout the past two hundred years of the monetary financial system. Since the start of the 19th Century, the stock markets of United States and United Kingdom have seen 12 (US) and 11 (UK) stock market crashes, around half of these have been crashes of greater than 40%. The most severe crisis in both markets was the 1929 Wall Street Crash in which the US stock market crashed 85% and the UK stock market crashed 72%. Figure 1.1 shows the historical severity of crashes on the US stock market.

Figure 1.1: Historical Severity of Market Crashes on US Stock Market US Stock Market Crashes

1845 Railway Mania Bubble UK 1997 Asian Crisis 1866 Collapse of Overend and Gurney 1825 Latin American Crisis 1983 Latin American Debt Crisis 1837 Cotton Crisis 1857 Railroad Mania Bubble US 1907 Knickerbocker 1987 Black Monday 2001 Dotcom 1893 Baring Bank Crisis 1973 Oil Crisis 1873 Long Depression 2008 Great Financial Crisis 1929 Wall Street Crash 0% 20% 40% 60% 80% 100% Stock Market Crash Peak to Trough Source: Historical Casebook of Financial Crises. Cambridge Centre for Risk Studies. In pre-publication.

1 Publications from the Cambridge Centre for Risk Studies research programme into financial catastrophe risk can be found at http://cambridgeriskframework.com/downloads

Cambridge Centre for Risk Studies 3 Cambridge Global Risk Outlook 2018 Threat Profiles

Historically each country has had its own market, and markets have behaved independently – each having its own crises periodically, but over time these national markets have become increasingly connected, and behaves as an international financial system of aligned national markets. We model the potential for market crashes with each country having its own national market, but we categorize each national market by its degree of connectedness into the global financial system.

The International Monetary Fund categorizes the various free market national economies and their central banks into ‘core’ markets, that are highly interconnected with each other at the centre of the global financial system, and ‘periphery’ markets that are linked to the core, sometimes through historical links to some of the national markets in the core, but that behave more independently of the core system. Centrally planned economies are less reliant on private capital, and less connected to or affected by the international financial system. These are assumed to have their own periodic crises, but independent of the crises that may occur in the global financial system.

Figure 1.2 Structure of the global financial system, identifying ‘core’ national markets and ‘periphery’ national markets in regional clusters.

Source: A network analysis of global banking; Minoiu, Camelia ; Reyes, Javier A., IMF Working Paper. URL: http://www.imf.org/external/pubs/ft/wp/2011/wp1174.pdf

Market crashes can be triggered by a number of causes. The most common causes are asset bubbles: a sudden re-evaluation of the fundamental values of an asset class;2 and bank runs: the loss of public

2 The potential impact of asset bubbles as a cause of a financial crisis, using a potential property bubble as an example is explored in ‘Global Property Crash Stress Test Scenario’ published by Cambridge Centre for Risk Studies in 2016.

Cambridge Centre for Risk Studies 4 Cambridge Global Risk Outlook 2018 Threat Profiles

confidence that a financial institution can fulfil its obligations. Other potential causes include inflationary shocks;3 reserve currency shifts;4 fraud and rogue trader crises; flash crashes from automated trading systems; cyber crimes and cryptocurrency failures; and sudden changes in sentiment by the investment community.5 Two other potential causes of financial crises are modelled separately in the Cambridge Global Risk Model – a sovereign crisis, and a commodity price shock – because the characteristics of these events cause different types of consequences to the economy and have different geographical impacts and effects on the various sectors of the economy. Collectively, these various types of causes of financial crises are termed ‘endogenous processes’ because they arise from within the financial system, triggering failures of confidence or changed evaluations of fundamental value systems. In many financial crises, there are multiple causes and interlinked failures of confidence. The modelling of Market Crash in the Cambridge Global Risk Model is of endogenous financial crises. However, a small proportion of financial crises can also be caused by ‘exogenous’ factors – events such as other crises that impact the functioning of the financial system and the confidence of the traders and decision-makers within it. Historically, events such as major interstate conflicts, severe failures of food supply, and social unrest, have also triggered financial crises. Some estimates suggest that around 10% of historical financial crises are from exogenous causes.6 In the Cambridge Global Risk Model we model each individual threat type separately, and the consequences of the financial shock from this exogenous cause are included in the effects of the primary threat. Future versions of the model may explicitly model one threat scenario triggering another type of threat event in consequence, such as an interstate conflict triggering a market crash. In the model of Market Crash in the CGRM we represent only the endogenous causes of shocks coming from within the financial system itself. Each national market has historically experienced its own set of financial crises, for example a run on a bank in that country, or a bubble in valuing assets traded in that market. The frequency and severity of crises occurring in each market is derived from the historical rate of crises seen in the past. There is evidence that the frequency of market crashes may have increased with globalization, as the international financial system experiences shocks from each of its constituent national markets, and occasionally these spread from one market to another, through contagion.

Contagion occurs when one financial institution suffers distress and through its actions, causes other institutions financial difficulties. Contagion mechanisms include interbank lending (a bank recalls funds that it has lent to another bank, causing that bank to call in its own loans to other institutions); asset fire sales (when an institution that is short of funds sells off some of its investment assets, reducing the price of the asset type, causing devaluation of the investment portfolios of any other institutions that contain that asset type); cross-shareholding; and roll-over risk. Interaction between these mechanisms is more important than a single mechanism on its own.7

1.2 Mapping the Threat We categorize each country’s financial market by its level of connectivity, using:

- IMF categorization of core-periphery status (figure 1.2) - Volatility of local stock markets and correlation of returns to global markets8

Each country is then given a threat assessment grade (TAG) based on this data. All cities in the same country receive the same TAG as they are assumed to experience the same economic shock following a market crash.

3 Inflationary financial crises are explored in ‘High Inflation World Financial Catastrophe Stress Test’ published by Cambridge Centre for Risk Studies in 2016. 4 The potential disruption that could occur from the global financial system switching from the dollar to the renminbi as the standard reserve currency is described in ‘Dollar Deposed: Financial Catastrophe Stress Test’ published by Cambridge Centre for Risk Studies in 2016. 5 The potential for shocks resulting from divestment from the carbon economy is explored in ‘Unhedgeable Risk: How Climate Change Sentiment Impacts Investment’ published by Cambridge Centre for Risk Studies in 2016. 6 Cutler, Poterba, Summers (1989). 7 Caccioli, Fabio, Farmer, J. Doyne, Foti, Nick and Rockmore, Daniel, (2015), Overlapping portfolios, contagion, and financial stability, Journal of Economic Dynamics and Control, 51, issue C, p. 50-63. 8 CCRS analysis, Bloomberg L.P. (2017) MSCI Country Indices Price and Volatility History

Cambridge Centre for Risk Studies 5 Cambridge Global Risk Outlook 2018 Threat Profiles

Threat Assessment Grade No. of Examples Countries A Peripheral to International Financial System Afghanistan, Croatia, 72 but Highly Affected Tunisia B Local Markets Volatile - Influenced by Belgium, Russia, Sweden 22 International Financial System C Integral part of International Financial Canada, United Kingdom, 12 System - Stable, interlinked markets United States D Isolated from International Financial System 1 Cuba

1.3 Local Impact Severity Definitions Each city is analysed for the GDP impact and likelihood of experiencing the following characteristic market crash scenarios. A city is assumed to experience the market conditions of its national market.

Representative LIS Description From To Severity Level MC1 Stockmarket Index drops (peak to trough) by 10% in a 10% 5% 20% single quarter (e.g. Asian Crisis 1997) MC2 Stockmarket Index drops 50% (peak to trough) in a single 50% 20% 70% year (e.g. SubPrime 2008) MC3 Stockmarket Index drops 85% in a single quarter (e.g. 85% >70% Wall Street Crash 1929)

1.4 Quantifying the Threat We quantify the risk of a market crash of each local impact severity level for each threat assessment grade. These estimates are based off historical analysis of stock market crashes, such as the US financial crises presented in Figure 1.1. The return periods are adjusted from their historical baseline according to The Centre for Risk Studies’ research on the existing state of financial stability and regulation.

Annual Likelihood Return Period Threat Assessment Grade MC1 MC2 MC3 MC1 MC2 MC3 A Peripheral to International 0.24 0.024 0.01 4 42 100 Financial System but Highly Affected B Local Markets Volatile - Influenced 0.06 0.008 0.003 17 125 300 by International Financial System C Integral part of International Financial System - Stable, interlinked 0.025 0.004 0.002 40 250 500 markets D Isolated from International 0 0 0 Financial System

1.5 Vulnerability assessment We estimate the vulnerability of a city’s economy to a market crash is dependent on its reliance on private capital, i.e. the need to raise funding in external financial markets rather than through fiscal means. We use a country’s sovereign rating from S&P9 as a proxy to measure its level of private capital reliance, assuming countries that have higher debt ratings have higher access to capital markets and therefore would utilize that capacity.

9 S&P Global Sovereign Ratings List; URL: https://www.capitaliq.com/CIQDotNet/CreditResearch/RenderArticle.aspx?articleId=1894456&SctArtId=433731&from= CM&nsl_code=LIME&sourceObjectId=10199471&sourceRevId=1&fee_ind=N&exp_date=20270804-21:15:37

Cambridge Centre for Risk Studies 6 Cambridge Global Risk Outlook 2018 Threat Profiles

While countries which are more reliant on private capital have better debt ratings and therefore are less likely to default on payments, in the case of a financial crash, we expect these countries’ economies to be more vulnerable to market fluctuations.

This analysis is largely based off the historical case studies conducted by Cambridge Centre for Risk Studies.10

Vulnerability Assessment S&P Rating No. of Examples Grade Range Countries 5 High Reliance on Private AA+ to AAA 14 Germany, United Kingdom, Capital United States 4 Moderately High Reliance A- to AA 16 Belgium, Chile, Qatar on Private Capital 3 Moderate Reliance on BBB- to BBB+ 23 Brazil, Mexico, Russia Private Capital 2 Low Reliance on Private B- to BB+ 33 Argentina, Ecuador, Pakistan Capital 1 Very Low Reliance on Ungraded to 21 Afghanistan, Chad, Nepal Private Capital CCC+

1.6 Consequence Analysis The initial impact to GDP due to characteristic financial crash events is determined from prior in- depth scenario studies and subject matter expertise (see Global Property Crash Financial Catastrophe Stress Test, Dollar Deposed Financial Catastrophe Stress Test and High Inflation World Financial Catastrophe Stress Test at cambridgeriskframwork.com).

The subsequent recovery from the initial GDP shock is determined by the socioeconomic resilience of each city. Characteristic recovery profiles for each resilience level were determined also from prior scenario studies and subject matter expertise.

1.7 References A network analysis of global banking; Minoiu, Camelia ; Reyes, Javier A., IMF Working Paper. URL: http://www.imf.org/external/pubs/ft/wp/2011/wp1174.pdf

Global Property Crash Financial Catastrophe Stress Test, URL: http://cambridgeriskframework.com/getdocument/30

Dollar Deposed Financial Catastrophe Stress Test, URL: http://cambridgeriskframework.com/getdocument/32

High Inflation World Financial Catastrophe Stress Test URL: http://cambridgeriskframework.com/getdocument/31

Cutler, Poterba, Summers, (1988) What Moves Stock Prices? National Bureau of Economic Research Working Paper Series, Vol. 2538. URL: http://www.nber.org/papers/w2538.pdf

Unhedgeable risk: How climate change sentiment impacts investment (CISL, 2015) https://www.cisl.cam.ac.uk/publications/publication-pdfs/unhedgeable-risk.pdf

Caccioli, Fabio, Farmer, J. Doyne, Foti, Nick and Rockmore, Daniel, (2015), Overlapping portfolios, contagion, and financial stability, Journal of Economic Dynamics and Control, 51, issue C, p. 50-63.

10 Historical Casebook of Financial Crises. Cambridge Centre for Risk Studies. Pre-publication stage.

Cambridge Centre for Risk Studies 7 Cambridge Global Risk Outlook 2018 Threat Profiles

2 Sovereign Crisis

2.1 Threat Description There have been 120 sovereign defaults in the past 100 years, equating to more than one default a year on average. A country’s default on its borrowings is therefore not unlikely, although significant economic consequences are associated with the events both leading up to the default and afterwards. Impacts include inflation, lowered consumer spending and reduction in international trade, as was seen in Argentina’s partial default in 2014.

The main threat in a sovereign crisis is a cascade of defaults in which multiple countries default under similar conditions or from follow-on consequences. A cascade involving 4 or more countries has occurred on average every 14 years. The size of the economy defaulting is a key determinant. Trading flows between countries and intergovernmental loans are also critical as a single default can lead to financial burden on other countries.

Borrowers (sovereign governments) and foreign lenders (investors) face frictions that interact in a vicious circle:

- Creditors become more risk averse after any default - Any country default increases the cost of borrowing to other countries, particularly those with sub-optimal credit ratings - Foreign lenders have regulatory collateral constraints that limit their investment leverage in sovereign debt

In the 2015 Grexit crisis in which speculation heightened about Greece’s exit from the EU surrounding bailout negotiations, other sovereigns were also affected. Portugal sovereign debt CDS spreads implied default probability of 7.5% and Ireland around 5.3%.

Contagion ‘spillover’ occurs if creditors to defaulting country incur so much loss that they, in turn, become insolvent. However, credit markets in general perceive little risk of contagion from spillovers following a sovereign default.

2.2 Mapping the Threat We categorize the threat assessment group of each country using a blend of credit rating assessments from each of the principal rating agencies: Moody's; S&P; and Fitch.

Cambridge Centre for Risk Studies 8 Cambridge Global Risk Outlook 2018 Threat Profiles

Source: S&P Global Sovereign Ratings List; URL: https://www.capitaliq.com/CIQDotNet/CreditResearch/RenderArticle.aspx? articleId=1894456&SctArtId=433731&from=CM&nsl_code=LIME&sourceO bjectId=10199471&sourceRevId=1&fee_ind=N&exp_date=20270804- 21:15:37 Source: Bloomberg Terminal The subsequent categorization into was derived primarily from S&P ratings:

No. of Threat Assessment Grade Examples Countries A - Substantial Chance of Sovereign Default 37 Afghanistan, Sudan, Tunisia B - Significant Chance of Sovereign Default 41 Argentina, Lebanon, Vietnam C - Moderate Chance of Sovereign Default 28 China, Italy, Thailand D - Low Chance of Sovereign Default 15 Austria, Qatar, United Kingdom E - Very Low Chance of Sovereign Default 12 Australia, Germany, Sweden

2.3 Local Impact Severity Description Each city is analysed for the GDP impact and likelihood of experiencing a sovereign default scenario.

LIS Description SD1 Country defaults and reschedules its debt, devalues its currency substantially (e.g. 50%). Investors flee; FDI Lost

Cambridge Centre for Risk Studies 9 Cambridge Global Risk Outlook 2018 Threat Profiles

2.4 Quantifying the Threat For each threat assessment category, the likelihood of a sovereign crisis is determined by an historical analysis of defaults of the countries assigned to each credit rating group, i.e. threat assessment grade. A subject matter expert further adjusted these annual likelihoods based on whether the rate of defaults for each sovereign rating is below or above the historical baseline.

Historical Historical S&P No. of Default default Estimated No. of Defaults Rate per rate Annual Countries in past Country per within Probability 104 yrs year 10 yrs A - Substantial Chance of 37 34 0.92% 51.65% 1.400% Sovereign Default B - Significant Chance of 41 18 0.44% 28.70% 0.660% Sovereign Default C - Moderate Chance of 28 14 0.50% 4.88% 0.100% Sovereign Default D - Low Chance of 15 16 1.07% 0.86% 0.018% Sovereign Default E - Very Low Chance of 12 1 0.08% 0.78% 0.016% Sovereign Default

2.5 Vulnerability Assessment For sovereign crises, we have assumed one level of vulnerability across all cities and countries. The implication is that all cities incur the same percentage shock to GDP in the case of default. The differentiation between cities is in their recovery process in the years following a default.

2.6 Consequence Analysis The initial impact to GDP due to a sovereign default is determined from a combination of historical studies of GDP impacts following default as well as subject matter expertise.

The subsequent recovery from the initial GDP shock is determined by the socioeconomic resilience of each city. Characteristic recovery profiles for each resilience level were also determined from historical studies and subject matter expertise.

2.7 References Allen, F.; Gale, D. (2009). Understanding Financial Crises. USA: Oxford University Press.

Reinhart, Carmen M.; Rogoff, Kenneth S. (2010). This Time Is Different: Eight Centuries of Financial Folly. Princeton University Press.

Neal, L.; Coffman, D’Maris. (2014). A History of Financial Crisis. Routledge.

Turner, J.D. (2014). Banking in Crisis: the Rise and Fall of British Banking Stability, 1800 to the Present. Cambridge University Press.

Caccioli , Fabio; Farmer, J.D.; Foti, N.; Rockmore, D. (2013). How interbank lending amplifies overlapping portfolio contagion: A case study of the Austrian banking network. Papers 1306.3704, arXiv.org. https://ideas.repec.org/p/arx/papers/1306.3704.html

S&P Global Ratings. URL: https://www.spratings.com/sri/

Cambridge Centre for Risk Studies 10 Cambridge Global Risk Outlook 2018 Threat Profiles

3 Commodity Price Shock

3.1 Threat Description Sudden hikes in the prices of key commodities have historically been very damaging to economies, particularly those that are net commodity importers. Increased prices can lead to a reduction in production as well as a reduction in demand. We model this risk principally by the potential for unit energy prices to escalate dramatically. Energy is a key component of the pricing of all basic commodities, as fuel costs for transporting materials to market is a major cost. Price hikes in raw material commodities, consumer indices, and food, all correlate closely to price hikes in oil. Oil pricing is a systemic issue – it has a global price so a price hike is felt in every part of the world simultaneously, but it affects the economies of different countries and cities in individual ways. We map the dependency of the economy of each city on the ‘energy intensiveness’ of the economy of the country, i.e. the amount of GDP produced per unit of energy consumed. Some countries are more resilient than others. Countries that are self-sufficient in oil, or net producers, are less vulnerable than economies that are net importers, so importers are categorized separately from exporters. Note that the probability of oil price hikes is the same globally – all countries and cities have the same probabilities of experiencing the price shocks applied. The vulnerability of the economy to the price shock is the principal analysis. The world is currently dealing with a state of energy pricing that is in flux, that is well below the long term norm of the past several decades, and has defied the established wisdom of the previous regime of pricing. Over the past several decades, oil pricing was largely managed by an OPEC ‘swing’ control system: when market prices became too high, OPEC production would increase to countermand it, and when pricing dropped, OPEC production would reduce. In 2014, oil prices plummeted to unprecedented lows, ushering in a new era where production of shale oil and gas and wider-spread renewables have significantly weakened the ability of OPEC to control pricing fluctuations. In this new era, the pricing fluctuations over the last several decades are unlikely to be a good guide to the likelihood of future oil price shocks. Instead, we take a longer-term perspective, looking at how often new ‘regimes’ of pricing have replaced old ones: How the era of very low pricing of the 1960s was replaced by a new regulated high-price regime following the oil crisis of 1974, and when other supply and demand crises have occurred. Geopolitical crises can be drivers of these: the Iraqi burning of Kuwaiti oil fields in 1991 saw the price of oil jump by 60% in just over 10 weeks. In each regime shift, the economies of each country have gradually adjusted to the new reality, with winners and losers, but the initial shock has been highly damaging to the local economies of highly oil-dependent societies. We monitor the views of energy pricing analysts and propose to adjust the probabilities of future pricing shocks each year depending on market outlooks and projection data. The current market view is that the availability of shale and renewables is likely to keep global prices low and relatively stable for the foreseeable future. We currently parameterize the likelihood of future severe price shocks as being below the medium-term average that prevailed from 1974 to 2014.

3.2 Mapping the Threat Because the effect of an oil price hike is global, i.e. oil is a global commodity with a global price, the likelihood of an event is the same for each country and city. The vulnerability of each country and city’s economy to an increase in crude oil price is the principal analysis.

3.3 Scenario Local Impact Severity Description The commodity price shock scenario is a sudden increase in the unit spot price of crude oil (e.g. Brent ICE or West Texas Intermediate) over a single quarter. The shocks are calibrated to crude oil prices at the time of analysis, which for November 2017 was approximately $50 per barrel. Following the shock to oil prices, there is an additional pass-through impact to other commodities such as food items, industrial metals and chemicals.

Cambridge Centre for Risk Studies 11 Cambridge Global Risk Outlook 2018 Threat Profiles

Each city is analysed for the GDP impact and likelihood of experiencing the following characteristic commodity price shock scenarios.

LIS Description Representative From To value ($/bbl) ($/bbl) OP1 Sudden (i.e. within a single quarter) increase in unit 50% $75 $125 price of crude oil from $50/bbl to $100/bbl (100%); pass-through impacts seen in other commodities OP2 Sudden (i.e. within a single quarter) increase in unit 200% $125 $250 price of crude oil from $50/bbl to $150/bbl (200%); pass-through impacts seen in other commodities OP3 Sudden (i.e. within a single quarter) increase in unit 500% >$250 >$250 price of crude oil from $50/bbl to $300/bbl (500%); pass-through impacts seen in other commodities

3.4 Quantifying the Threat The likelihood of each scenario was determined from the empirical probability of monthly oil price returns from World Bank price data. The probability of each scenario captures the range of prices defined in the local impact severities. As price shocks are global, there is no geographic differentiation in likelihood of events.

Source data: CCRS analysis, World Bank Monthly Commodity Markets Prices http://pubdocs.worldbank.org/en/561011486076393416/CMO-Historical- Data-Monthly.xlsx A subject matter specialist further judges whether oil price returns given the current commodity price landscape would follow or deviate from the historical mean in the outlook period (three years). Note that since oil prices have not historically reached the level of OP3 severity, the estimated probability was determined through expert judgement.

Historical Annual Estimated Return Period Likelihood Return Period OP1 20 0.025 40 OP2 60 0.008333 120 OP3 Unknown 0.001 1000

Cambridge Centre for Risk Studies 12 Cambridge Global Risk Outlook 2018 Threat Profiles

3.5 Vulnerability Assessment The vulnerability assessment of a commodity price shock was done at a country level, assuming energy resources are largely managed on a national basis. The vulnerability of each country is determined by its reliance on fossil fuel per unit of GDP. This is estimated by:

Unit of Imported Fossil Fuel per unit of GDP = Energy Intensity * Fossil fuel energy consumption (% of total energy) * % of energy imported

Energy intensity: Units of energy (all types) per unit of GDP

- Description: GDP per unit of energy use (constant 2011 PPP $ per kg of oil equivalent) - Source: http://data.worldbank.org/indicator/EG.GDP.PUSE.KO.PP.KD/countries

Fossil fuel energy consumption: (% of total)

- Description: Fossil fuel comprises coal, oil, petroleum, and natural gas products. - Source: http://data.worldbank.org/indicator/EG.USE.COMM.FO.ZS/countries

Energy imports, net: (% of energy use)

- Description: Net energy imports are estimated as energy use less production, both measured in oil equivalents. A negative value indicates that the country is a net exporter. Energy use refers to use of primary energy before transformation to other end-use fuels, which is equal to indigenous production plus imports and stock changes, minus exports and fuels supplied to ships and aircraft engaged in international transport. - Source: http://data.worldbank.org/indicator/EG.IMP.CONS.ZS/countries

Note that future metrics to determine vulnerability to a commodity price shock might include energy use from alternative sources such as solar, wind or nuclear energy. Further, the emergence of a “shale revolution”, also known as fracking, is supposed to have changed the energy landscape11; there is now a considerable new lever in influencing oil prices as control of oil production shifts away from OPEC. The resulting metric of fossil fuel reliance was used to categorize each country to the following threat assessment grades:

Vulnerability Assessment Units of Fossil Fuel No. of Examples Grade Imported per Unit Countries GDP (kg oil equiv./$) 1 High Vulnerability of 5-15 21 Armenia, Economy to Oil Price Shock Lebanon, Turkey 2 Moderate Vulnerability of 2-5 25 Austria, India, Economy to Oil Price Shock United States 3 Some Vulnerability of 0-5 28 Argentina, China, Economy to Oil Price Shock Sweden 4 Exports marginally more oil N/A 9 Cameroon, Peru than it consumes - economy would marginally benefit from oil price increase 5 Exports significantly more N/A 10 Australia, Russia oil than it consumes - economy would benefit from oil price increase 6 Exports a lot more oil than it N/A 14 Iraq, Saudi Arabia consumes - Economy would see significant benefit from oil price increase

11 Manning, R. 2014. The Shale Revolution and the New Geopolitics of Energy. https://www.files.ethz.ch/isn/185485/Shale_Revolution_and_the_New_Geopolitics_of_Energy.pdf

Cambridge Centre for Risk Studies 13 Cambridge Global Risk Outlook 2018 Threat Profiles

The countries which export more oil than it consumes is expected to see a net positive impact to GDP in the case of an oil price increase. Given the Cambridge Risk Framework currently only captures negative shocks to GDP, countries which are classified in categories X, Y, Z will incur zero impact.

3.6 Consequence analysis It is important to bear in mind that not all oil price shocks are the same.

There are several reasons why oil price shocks are fundamentally different from the increases in other goods.

1. Energy prices at times experience sustained increases not typical of other goods and services 2. Energy prices matter more because they are demand-inelastic 3. Energy price fluctuations are deemed exogenous to most national economies 4. Historically, large increases in energy prices have typically been followed by recessions

Most oil price shocks are driven by a combination of strong world-wide demand for crude oil, and because of the close link between economic activity and oil consumption this usually coincides with periods of strong economic growth. This has the effect of shifting the precautionary demand for crude oil representing the market’s perception of the likelihood of a future shortfall in the supply of oil. The likelihood of a shortfall in crude production is driven by expectations about future demand for oil as well as expectations on the ability of producers to meet these expectations.

A modern view of oil price shocks is that oil price shocks affect the economy primarily through their effect on consumer and firm expenditures. In this view, higher energy prices cause both a reduction in aggregate demand and a shift in expenditures which causes a ripple effect through the economy. It is also important to note the high oil prices do not occur in isolation but are driven by demand and supply shocks that each have different macro-economic impacts.

Most forecasters, supported by the International Energy Agency (IEA) predicts that every $10 increase in crude oil price trims global growth by 0.5%. 12

Source: http://voxeu.org/article/oil-shocks-around-world-are-they-really-bad

12 International Energy Agency, Analysis of the Impact of High Oil Prices on the Global Economy, May, 2004. https://www.iea.org/textbase/npsum/high_oil04sum.pdf

Cambridge Centre for Risk Studies 14 Cambridge Global Risk Outlook 2018 Threat Profiles

3.7 References World Bank Monthly Commodity Markets Prices http://pubdocs.worldbank.org/en/561011486076393416/CMO-Historical-Data-Monthly.xlsx

http://voxeu.org/article/oil-shocks-around-world-are-they-really-bad

International Energy Agency, Analysis of the Impact of High Oil Prices on the Global Economy, May, 2004. https://www.iea.org/textbase/npsum/high_oil04sum.pdf

Manning, R. 2014. The Shale Revolution and the New Geopolitics of Energy. https://www.files.ethz.ch/isn/185485/Shale_Revolution_and_the_New_Geopolitics_of_Energy.pdf

Blanchard, O.J., Gali, J. (2007) The Macroeconomic Effects of Oil Shocks: Why are the 2000s So Different from the 1970s? NBER Working Paper No. 13368. URL: http://www.nber.org/papers/w13368

Cambridge Centre for Risk Studies 15 Cambridge Global Risk Outlook 2018 Threat Profiles

Part B Geopolitics and Security

4 Interstate Conflict

4.1 Threat Description Military conflicts have historically caused major economic losses to the protagonists and to the trading activities of the world and financial health of unaffected parties. Armed conflicts litter the records of history.

Rank Deaths Conflict Dates

1 40 - 72 million World War II 1939-1945

2 15 - 65 million World War I 1914-1918

3 5 - 9 million Russian Civil War 1917-1921

4 2.5 - 5.4 million Second Congo War 1998-2003

Afghan Internal 5 1.5-2 million 1979-1984 War

6 1-2 million Sudanese Civil War 1983-2005

7 1-3 million Nigerian Civil War 1967-1970

8 800k - 3 million Vietnam War 1955-1975

Soviet War 9 600k - 2 million 1980-1988 Afghanistan

10 500k - 2 million Iran–Iraq War 1980-1988

11 500k - 2 million Mexican Revolution 1911-1920

12 400k – 4 million Korean War 1950-1953

Table 4.1: Major conflicts in the 20th Century, ranked by number of deaths caused. Table 4.1 shows 12 conflicts in the past century that have each caused millions of deaths. Even the facts about these events are uncertain because of the great disruption and chaos that they cause. The 20th century was one of the bloodiest for conflicts, but by no means unprecedented. Studies of wars since 500 B.C. show that the most serious wars and atrocities – those that killed more than a tenth of a percent of the population of the world, have been pretty evenly distributed through the past 2,500 years of history13. However, the threat of war is not high on the risk assessment of many people. In surveys of perception of risks by industry and political leaders, conflict between nations tends to be low down the risk rankings14. The Long Peace

The reason for this is that the period since the end of the second world war has been a lengthy period of peace, with no conflicts between major military powers, despite (or perhaps because of) major powers possessing nuclear weaponry that could inflict death tolls much higher than those achieved with conventional weaponry. This is known by political historians as ‘The Long Peace’. It saw the world change from a cold war face-off between US and USSR, in which many commentators saw

13 Pinker (2011) p238. 14 In the 2014 Global Risk Perception Report derived from surveys of risk perception by more than a thousand participants, interstate conflict does not make the top ten, and is ranked below average in impact. (WEF 2014).

Cambridge Centre for Risk Studies 16 Cambridge Global Risk Outlook 2018 Threat Profiles

nuclear war as inevitable, to one of a single military super-power, the United States, policing a ‘Pax Americana’. Many reasons have been cited for this, including the extension of democracy (democratic states rarely if ever go to war with each other); globalization and the inter-dependency of trade (it is economically too costly to go to war); education of the population and growing intolerance of political belligerence by their leadership; increasing acceptance of international law and more influential United Nations institutions that maintain peace; and the decreasing ‘business case for war’ in terms of gaining advantage from conflict. Changing nature of modern war

And yet most nation states retain their capability for war, and the world still devotes 2.5% of its GDP on military expenditure15. Military spending worldwide today is at similar levels to the height of the cold war in 198816, after a dip to the mid-1990s. Modest reductions in military spending by the West (mainly Europe) have been offset by rises in spending in the developing markets, Eastern Europe, and Russia. Nation states do not seem to assume that the threat of war has diminished, and they continue to see value in military expenditure to protect themselves against it. The Long Peace period has not been conflict free. In fact records suggest that there may have been over a million deaths in at least 700 militarized conflicts in the past 25 years17. However the nature of conflict appears to have shifted during this period of the Long Peace, to one of insurgency, asymmetrical warfare, civil war, and low-grade political violence, rather than interstate military confrontation between major powers. Political scientists are split between those who believe that violence is decreasing generally across society and argue that war is a permanently reduced threat18, and those who believe that a half century of low conflict activity is a phase that could end at any time19. Conflict theory suggests that underlying tensions rise and fall with shifts in the balance of power, and disputes arise that provide the conditions for a potential conflict, but that common sense often prevails and the cost of a prospective war is usually sufficient to force antagonists to come to a peaceful resolution. Overconfidence and ‘positive illusions’ can trigger wars: “opponents rarely go to war without thinking they can win, and clearly one side must be wrong – this conundrum lies at the heart of the ‘War Puzzle’: rational states should agree on their differences in power and thus not fight”20. Random chance may also account for the difference between a tense situation escalating into a conflict, rather than defusing through negotiation. Statistical analysis of intervals between wars (and the duration of wars) suggests that starts and ends of wars are consistent with observations of random processes21 – wars often start by accident. The Long Peace could be as much about chance not having given rise to a war trigger event as there being a reduction in the underlying tensions and motivations for war. Other theories suggest that periods of hegemony – one dominant military power – are associated with periods of peace, but that when the world order is threatened, such as a new challenger to the superpower, the likelihood for conflict increases. In this interpretation, the United States dominance that has ensured a half-century of low activity rates for conflict, could be entering a new period of challenge from China and possibly from a resurgent Russia. The chance of a conflict may be increasing again as regional powers seek to challenge and define a new world order. Statisticians have observed from the statistics of deadly quarrels22, that like many types of catastrophes, the ranked distribution of size of wars conforms to a power law – there are a lot of small wars, and only a few large wars and their ratio follows a logarithmic progression.

Preparing for the possibility of a war remains a challenge for business managers. It is certainly difficult to imagine the end of the prosperous peace that we have come to expect. It is difficult for

15 Stockholm International Peace Research Institute (2013). 16 SIPRI (2012). 17 Uppsala Conflict Database Program, University of Uppsala, from 1990. UCDP Conflict Encyclopedia. 18 A decreasingly violent society is argued by Pinker (2011) The Better Angels of Our Nature. 19 Sobek (2008). The Causes of War. Cambridge, UK. 20 Johnson (2004). Overconfidence and War: the havoc and glory of positive illusions. 21 Richardson (1960) shows that the timing of wars is ‘Poissonian’ – consistent with a random roll of the dice each year. 22 First observed by Lewis Fry Richardson in 1960.

Cambridge Centre for Risk Studies 17 Cambridge Global Risk Outlook 2018 Threat Profiles

rational people to imagine the occurrence of a war – and has always been: “In 1914, Europe sleepwalked into a war that no one expected.”23

4.2 Mapping the Threat

Military Power We categorize each country by its strength as a military power, based on military power index, produced by globalfirepower.com. Each country is categorized as one of the following.

World Power Examples Ranking 1 Very Minor >75 South Sudan, Congo, Cambodia

2 Minor Power 30 to 75 Ethiopia, Malaysia, Norway

3 Medium Power 15 to 30 Vietnam, Iran, Ukraine

4 Major Power 4 to 15 Japan, UK, India

5 Superpower 1 to 3 China, Russia, USA

Source: CCRS analysis, Global Firepower Index: https://www.globalfirepower.com/countries-listing.asp

A subject matter expert further provided judgement on individual country results from Global Firepower. An additional layer of subjective expertise is necessary to account for factors such as the current geopolitical landscape and alliances.

Magnitude of Conflict Through the development of a comprehensive historical catalogue, and detailed analysis of several examples a simple conflict magnitude scale has been derived to classify the various types of war. The scale is based upon a number of factors (e.g. size of belligerents, length of conflict, casualties, disruption, economic and social impact, etc.), and ranges from one to five. Conflict magnitudes are characterized by their bilateral pairing of the military power of their protagonists. We assume conflicts are initiated initially between two countries. Conflicts can of course escalate and draw in the allies of those countries to become a multilateral conflict between multiple powers, but for simplicity we assume that the conflict is constrained between two countries.

Typically conflicts arise between two powers of similar strength, or a stronger power goes to war with a weaker power.

Level 1 Conflict: Minor Power vs Minor Power (or Very Minor)

Level 1 Conflicts are relatively frequent and are generally limited in their social, economic and political impact at a regional and global level. They are limited in their modes of warfare and geographic scale and have primarily been fought for economic or political reasons, and are characterised by a roughly symmetrical, and relatively low, distribution of power between belligerents. While relatively minor countries in terms of size and strength, if they are producers of a particular commodity, there can still be significant ramifications for the global economy (e.g. the Iraqi burning of Kuwaiti oil fields saw the price of oil jump by 60% in just over 10 weeks). We currently identify [20] candidate bilateral pairings worldwide as potential level 1 conflicts.

Level 2 Conflict: Medium Power vs Minor Power

23 Clark (2013). The Sleepwalkers: How Europe Went to War in 1914.

Cambridge Centre for Risk Studies 18 Cambridge Global Risk Outlook 2018 Threat Profiles

Level 2 Conflicts occur fairly frequently and have more widespread economic, political and social implications than Level 1 Conflicts. Although such invasions are characterised by an asymmetrical distribution of power, they have also led to protracted conflicts through the application of insurgency/guerrilla style warfare. We currently identify [35] candidate bilateral pairings worldwide as potential level 2 conflicts.

Level 3 Conflict: Medium Power vs Medium Power

Level 3 conflicts involve the major industrialized nations of the world. Wars between two tier 2 countries have been driven primarily by economic and political factors in the 20th Century (e.g. Egypt-Israel & India-Pakistan). These wars are more likely to occur and continue during periods of bi- or multi-polarity, particularly when there is a vacuum in global leadership.

We currently identify [15] candidate bilateral pairings worldwide as potential level 3 conflicts.

Level 4: Superpower vs Minor Power

Superpowers carry out military interventions against minor powers that they perceive to threaten their interests, and where they are unlikely to trigger wider conflicts against client states of another superpower. Ideological and political motivations have been the source of the recent US-led invasions of Iraq and Afghanistan.

Level 4 conflicts of the past 50 years include the Vietnam war, the two Gulf wars, and Libya. These have typically involved consortiums of allied countries, acting under an international legal mandate, such as United Nations.

We currently identify [22] candidate bilateral pairings worldwide as potential level 4 conflicts.

Level 5 Conflict: Superpower vs Medium or Major Power

Level 5 conflicts are rare. Since the formation of the G20 there has yet to be a regional war between a superpower and a G20 nation. However, conflicts earlier in the 20th century, notably the Russo-Japanese War and the Second- Sino Japanese War have both been considered ‘Great Wars’ with significant global ramifications. The Second-Sino Japanese War was the longest conventional war of the 20th century and came at a huge human and economic cost: there were 15-20 million casualties and both economies lost tens of billions in due to physical damage and lost production.

The larger the powers involved, the more likely the conflict is to escalate and to draw in allies and multiple other protagonists. Most medium or major powers are highly allied to others, including other superpowers. Superpowers are extremely wary of triggering a conflict with another superpower. However, there is the potential for wars against proxies, allies and client states of other superpowers, and circumstances under which a superpower may consider it appropriate to conduct military operations against a medium power, judging it of low risk to provoke a rival superpower retaliation. Modern wars of this type are likely to engage in new forms of warfare, including cyber war, disinformation, and economic destabilization of each other’s economy. We currently identify [15] candidate bilateral pairings worldwide as potential level 5 conflicts

Level 6 Conflict: Superpower vs Superpower

We do not model level 6 conflicts. The three current global nuclear superpowers of United States, Russia, and China are extremely unlikely to go to war directly with each other, because of the scale and consequences of the world war that would result, and the potential for the exchange of thermonuclear weapons. The principles of Mutually Assured Destruction have acted as a key defense mechanism in enforcing the Long Peace for many decades. In the event that two superpowers did go to war, the analysis of economic impact and consequences would not be a useful risk management tool.

Identifying and prioritizing potential candidate conflict pairings

Cambridge Centre for Risk Studies 19 Cambridge Global Risk Outlook 2018 Threat Profiles

Each potential candidate conflict pairing of countries is derived from analysis of a number of factors, including any recent military skirmishes between countries, recent territorial, trading, or resources disputes, belligerent statements, historical wars between protagonists, religious or ethnic divides, and whether the two countries share a mutual border. Candidate pairs in each conflict level are ranked and prioritized each year according to the relative likelihood of this candidate scenario being the one that occurs in that conflict level.

Likelihood of conflict outbreak The likelihood of a conflict occurring in each Conflict Level is derived from the number of historical wars that there have been in each of the Conflict Levels since 1947 (the advent of the ‘Long Peace’), expressed as an annual probability. We assume that conflicts will continue to occur at the rate they have done since the end of the second world war. We do not include a future world war as a possibility in the modelling.

Number of Return Annual conflicts since Period Frequency 1947 1 Minor Power vs Minor Power 20 3.4 0.30 2 Medium Power vs Minor Power 23 2.9 0.34 3 Medium Power vs Medium Power 4 16.8 0.06 4 Superpower vs Minor Power 9 7.4 0.13 5 Superpower vs Medium or Major Power 2 33.5 0.03

Source: CCRS Analysis, Ghosn, Faten, and Scott Bennett. 2003. Codebook for the Dyadic Militarized Interstate Incident Data, Version 3.10. URL: http://correlatesofwar.org. The likelihood of occurrence of each Conflict Level is then distributed across the candidate conflicts according to their relative likelihood of occurrence to derive an annual likelihood for each individual candidate conflict.

For example, if there are 35 candidate scenarios for a Level 2 Medium Power vs Minor Power conflict, the most likely candidate is assigned a score of 10, and the least likely candidate is assigned a score of 1, with several candidates having the same score. The total scores sum to e.g. 157. The top candidate scenario is assigned 10/157 = 6.37% the likelihood. (the bottom candidate is assigned 1/157 = 0.64%). The annual likelihood of a Level 2 conflict is 0.34 (based on a rate of occurrence of 23 events in 70 years). We assume that the total list of 35 candidates captures around 90% of the likelihood (i.e. that we have identified 9 out of 10 of the future scenarios for a conflict of a Medium Power vs Minor Power). We distribute the annual probability according to each candidate’s share, so that the top candidate has an annual likelihood of occurrence of 0.017 (i.e. a ‘return period’ of 58.1 years).

Impact on Cities in Conflict Scenarios For the two countries in the candidate conflict scenario, we estimate the likelihood that cities will experience effects that will cause economic loss. These estimates are approximations representing the theory that cities in a country with weaker military power are more likely to suffer damages and losses while balanced powers would be equally affected.

In a conflict with If country is 1 - Very 2 - Minor 3 - Medium 4 - Major 5 - Superpower Minor Power Power Power 1 - Very Minor 50% 75% 90% 100% 100% 2 - Minor Power 25% 50% 75% 100% 100% 3 - Medium 10% 25% 50% 75% 100% Power 4 - Major Power 0% 0% 25% 50% 90%

5 - Superpower 0% 0% 0% 1%

Cambridge Centre for Risk Studies 20 Cambridge Global Risk Outlook 2018 Threat Profiles

The ranking of cities currently engaged or potentially will engage in a conflict along with its likelihood of being impacted are then categorized into threat assessment grade categories.

Threat Assessment Grade No. of Example Countries Countries A Very High Threat of Interstate Conflict 7 Afghanistan, Iraq Impacting Cities in Homeland B High Threat of Interstate Conflict 7 Armenia, Egypt Impacting Cities in Homeland C Moderate Threat of Interstate Conflict 16 Japan, Pakistan, Philippines Impacting Cities in Homeland D Low Threat of Interstate Conflict 18 Cambodia, India, Argentina Impacting Cities in Homeland E Major Power with Some Threat of 3 China, United States Interstate Conflict Impacting Cities in Homeland F Major Power with Very Low Threat of 2 France, United Kingdom Interstate Conflict Impacting Cities in Homeland G Conflict is Possible but No Specific 54 Bulgaria, Singapore, South Scenarios Identified Africa

4.3 Local Impact Severity Each city is analysed for the GDP impact and likelihood of experiencing the following characteristic interstate war scenarios.

LIS Description

IW1 City mobilized for war, but not attacked; industrial activity switches from commercial to military production; consumer demand and investor confidence drops, some outmigration; conflict impact lasts for 1 year

IW2 City suffers sporadic attack from occasional missiles or aerial bombardment (and possible damage from cyber attack); city is mobilized for war; consumer demand and investor confidence drops, large outmigration; conflict impact lasts for 2 year

IW3 City is the target of strategic bombing by enemy forces, destroying industrial and commercial output and military facilities in the city; significant outmigration. Possible rebuilding afterwards by major injection of capital; conflict impact lasts for 3 years

4.4 Quantifying the Threat The likelihood that a city of each threat assessment is impacted by an interstate conflict for each local impact severity was based off the overall historical average of conflicts since 1947. A subject matter expert determined according the current geopolitical landscape if interstate conflict events are following or diverging from its historical mean.

Annual Likelihood Return period Threat Assessment Grade IC1 IC2 IC3 IC1 IC2 IC3 A Very High Threat of Interstate 0.062 0.0372 0.0186 16 27 54 Conflict Impacting Cities in Homeland B High Threat of Interstate Conflict 0.0372 0.0155 0.0062 27 65 161 Impacting Cities in Homeland

Cambridge Centre for Risk Studies 21 Cambridge Global Risk Outlook 2018 Threat Profiles

C Moderate Threat of Interstate 0.0186 0.0037 0.0019 54 269 538 Conflict Impacting Cities in Homeland D Low Threat of Interstate Conflict 0.0062 0.0021 0.0009 161 484 1076 Impacting Cities in Homeland E Major Power with Some Threat of 0 0.0003 0 3181 Interstate Conflict Impacting Cities in Homeland F Major Power with Very Low Threat of 0 0.0002 0 6362 Interstate Conflict Impacting Cities in Homeland G Conflict is Possible but No Specific 0 0 0 Scenarios Identified

4.5 Vulnerability Assessment All cities and countries are assumed to be equally vulnerable to an interstate conflict of a given magnitude (in terms of economic impact) in its first year. The underlying theory is that there are few characteristics of a city that could ameliorate the impact of being attacked once it has happened, particularly for IW2 and IW3 scenarios. Cities that are more resilient can recover more quickly however.

4.6 Consequence Analysis The initial impact to GDP due to an interstate conflict is determined from a combination of historical studies of GDP impacts following a war, scenario studies, as well as subject matter expertise. Please see China-Japan Geopolitical Conflict Scenario at http://cambridgeriskframework.com/downloads.

The subsequent recovery from the initial GDP shock is determined by the socioeconomic resilience of each city. Characteristic recovery profiles for each resilience level were also determined from historical and scenario studies as well as subject matter expertise.

4.7 References 2017 Military Strength Ranking (2017). URL: https://www.globalfirepower.com/countries-listing.asp

Ghosn, Faten, and Scott Bennett (2003). Codebook for the Dyadic Militarized Interstate Incident Data, Version 3.10. URL: http://correlatesofwar.org.

Richardson, Lewis. (1960) Statistics of Deadly Quarrels. Boxwood Press.

Uppsala Conflict Data Program (2017). Uppsala University UCDP Conflict Encyclopedia. URL: www.ucdp.uu.se.

Pinker, S. (2011). The Better Angels of Our Nature: A History of Violence and Humanity. London: Penguin Books.

Sobek (2008). The Causes of War. Cambridge, UK.

Johnson, Dominic D. P. (2004). Overconfidence and War: The Havoc and Glory of Positive Illusions.

Bowman, G.; Caccioli, F.; Coburn, A.W.; Kelly, S.; Ralph, D.; Ruffle, S.J.; Foulser-Piggott, R. (2014). Stress Test Scenario: China-Japan Conflict; Cambridge Risk Framework series; Centre for Risk Studies, University of Cambridge.

Cambridge Centre for Risk Studies 22 Cambridge Global Risk Outlook 2018 Threat Profiles

5 Terrorism

5.1 Threat Description Political violence through terrorism has a long history. The modern era of terrorism is characterised by militant Islamic groups, since the advent of Al Qaeda and its various factions, including Islamic State, Boko Haram, and others. Whereas occasional sporadic terrorist attacks may occur in any country, a sustained effective campaign of terrorism capable of causing prolonged economic harm would need to be driven by one of three major underlying factors fuelling political instability: Separatism One sector of the population is politically alienated from the majority population, and seeks to break away and have its own independent state where their human rights and economic aspirations would be better respected. Regime Change There is substantial popular support amongst the aggrieved and oppressed for a violent overthrow of the prevailing political system to effect regime change. Sectarian Conflict There is substantial support amongst devotees of a particular religion or sect to seek a violent change of the prevailing political system in favour of one centred more on their own religious principles. The creation of an Islamic state based on Sharia Law is an aspiration and prime motivation for Jihadi terrorism. A number of countries are economically stable against terrorist action in that there is no violent internal separatist movement or sectarian conflict; political change takes place via the ballot box rather than the bomb; and indigenous support for the creation of an Islamic state is feeble. Furthermore, taking advantage of widespread, massive and indiscriminate electronic surveillance and intelligence gathering to supplement the work of informants, proficient counter-terrorism and security forces are able to interdict the vast majority of terrorist plots. Accordingly, whilst terrorism risk remains persistent and universal, and there is a possibility of a spectacular economic loss to the civil aviation industry, the annual probability of a sustained terrorist campaign causing significant economic harm is essentially negligible for the countries of North America, Western Europe, Japan, Singapore, Australia, and New Zealand. Apart from the Muslim provinces of Russia and China (Xinjiang), this holds also for cities in Russia and China. The Cambridge Centre for Risk Studies has compiled case studies of terrorism impact on local economies.24 In Western economies such as Paris, London and New York, the short term impacts are primarily due to the ‘fear factor’ which dampens confidence and consumer spending, and secondly, the negative impact on the tourism and leisure sectors.

5.2 Mapping the Threat The Cambridge model of terrorism activity is derived from the Global Terrorism Index (GTI)25. GTI is an attempt to systematically rank the nations of the world according to terrorist activity. The index combines several factors associated with terrorist attacks to build a thorough picture of the impact of terrorism over a 10-year period, illustrating trends, and providing a data series for analysis by researchers and policymakers. It is the product of Institute for Economics and Peace (IEP) and is based on data from the Global Terrorism Database (GTD) which is collected and collated by the National Consortium for the Study of

24 A Risk Analysis Retrospective on the 2015 Paris Terrorist Attacks (http://cambridgeriskframework.com/getdocument/66)

25 Global Terrorism Index 2017. (2017). Institute for Economics and Peace. URL: http://visionofhumanity.org/app/uploads/2017/11/Global-Terrorism-Index-2017.pdf

Cambridge Centre for Risk Studies 23 Cambridge Global Risk Outlook 2018 Threat Profiles

Terrorism and Responses to Terrorism (START) at the University of Maryland. The GTD has codified over 104,000 cases of terrorism. Terrorism Threat Assessment Gradings for countries were assigned from GTI as follows: Threat Assessment Grade Approximate No. Examples GTI Rank of Range Cities

A Extremely High Terrorism Activity 1-2 2 Kabul, Baghdad B Very High Terrorism Activity 3-6 8 Lagos, Karachi, Sana'a C High Terrorism Activity 6-20 36 Tehran, Mumbai, Istanbul D Moderate Terrorism Activity 20-45 32 Tel Aviv, Riyadh, Moscow 45-70 17 Dublin, Mexico City, Kuala E Moderate Small-Scale Terrorism Lumpur 70-105 30 Sao Paulo, Managua, Caracas F Potential Small Scale Terrorism 2 Paris, London G Very High Value Target, Highly Defended 120 Shanghai, Los Angeles H High Value Target, Highly Defended 2 Washington DC, New York I Highest Value Target, Very Highly Defended J Low Terrorism Activity or No Data 30 Zagreb, Riga, Seoul

Terrorism risk is a matter of defensive capability in addition to terrorist intent. There are several countries where the risk of future terrorist attacks may not be captured by the historical incident rates of past attacks. These are principally the countries of the Western alliance against terror who feature highly on terrorist measures of intent to attack, but whose security and counter-terrorism activities are strong and manage to maintain a high interdiction rate against planned attacks. These are represented by grade G, H, and I. Terrorism attacks are acts of political violence, protest, and attempts to provoke media attention. For this reason, multi-perpetrator, ‘macro’ terrorist attacks are more commonly targeted on the major cities and political centres. This is represented by grade G, indicating a slightly higher likelihood of a macro-terror attack as these cities are very high value targets. Grade I represents cities which are the highest value targets and therefore have some likelihood of an attack using weapons of mass destruction, although small.

5.3 Local Impact Severities The definition of the three local impact severities for Terrorist Attacks representative of micro terror attacks, a macro terror attacks, and an unlikely but severe attack using weapons of mass destruction. Each city is analysed for the GDP impact and likelihood of experiencing the following characteristic terrorism scenarios. LIS Description TR1 Sustained campaign of small-scale terror attacks (e.g. individual and mass-attack shootings, poisonings, food chain sabotage etc.) over a period of 9 months causes fear and distrust in urban population, leading to loss of consumer confidence, demand drop and loss of external investment TR2 Coordinated series of simultaneous high profile large terrorist attacks (e.g. major truck bombings, airplanes into buildings or other surprise destructive events) causes horrific loss of life and major destruction to property in and around city centre, leading to several years of depressed econmic activity and reduced external investment

TR3 WMD Terrorist Attack - City is attacked by sophisticated terrorist operation using weapons of mass destruction; (e.g. anthrax, air-dispersed bio-weapons, or chemical or radioactive contaminant, or small yield nuclear detonation) kills large numbers of people and contaminates many buildings in Central Business District, requiring years of recovery

Cambridge Centre for Risk Studies 24 Cambridge Global Risk Outlook 2018 Threat Profiles

5.4 Quantifying the Threat26 The average annual probability over the coming decade of a micro (TR1) or macro (TR2) terrorist campaign is estimated by the following procedure. 1. Using the resources of the Global Terrorism Database, the statistics of terrorist attacks since 2002 are downloaded and examined for each country. 2. For each year since 2002, and for each country, an assessment is made as to whether the terrorist activity reached a sufficient level to be labelled a micro or macro campaign. 3. A base estimate of annual micro/macro campaign probability is made from the aggregate historical experience since 2002. Allowance is made for the chance that more plots might counterfactually have been successful. 4. This base estimate is updated by a political judgement as to the likelihood of political instability over the coming decade, and the extent to which the future decade will differ from the past decade. Migration of terrorist activity is an obvious prospect, illustrated by Lagos in Nigeria, which has recently become targeted by Boko Haram. 5. Probability figures are rounded to reflect the resolution of the results, and countries are ranked according to these figures. A relative risk comparison between countries is then made as a test of internal consistency and robustness. 6. The lower threshold for inclusion in the listings for micro and macro terrorism is 1%. Below this level, the risk of economic harm may be considered negligible, relative to other factors. A WMD terrorist attack (TR3) has been a serious western concern since 9/11. Given the high likelihood of plot interdiction for macro-terror plots involving a sizeable team of operatives, the most likely scenario is for new weapon capability to be deployed initially with a cautiously small or moderate payload. Stockpiling of chemical or biological weapon material would run the risk of arrest before deployment. A successful minor WMD attack would inevitably be followed by a massive security crackdown, which would greatly reduce the chance of a subsequent large weapon attack. The exception would be for a small nuclear detonation device. Since there is no miniaturized version, even the smallest nuclear bomb might well generate loss requiring a 3 year economic recovery period. A logic-tree analysis indicates a slight annual probability of about 1/5000, with the risk split between New York (0.0001) and Washington DC (0.0001).

Annual Likelihood Return Period Threat Assessment Grade TR1 TR2 TR3 TR1 TR2 TR3

A Extremely High Terrorism 0.3704 0.3704 0 3 3 Activity 0.1852 0.0463 0 5 22 B Very High Terrorism Activity 0.1204 0.0189 0 8 53 C High Terrorism Activity 0.0556 0.0056 0 18 180 D Moderate Terrorism Activity E Moderate Small-Scale 0.0278 0 0 36 Terrorism F Potential Small Scale 0.0056 0 0 180 Terrorism G Very High Value Target, 0.0019 0.0067 0 541 150 Highly Defended H High Value Target, Highly 0.0019 0.0005 0 541 2160 Defended I Highest Value Target, Very 0.0019 0.0033 0.0001 541 300 10000 Highly Defended J Low Terrorism Activity or No 0 0 0 Data

26 From the work of Gordon Woo for Centre for Risk Studies

Cambridge Centre for Risk Studies 25 Cambridge Global Risk Outlook 2018 Threat Profiles

5.5 Vulnerability Assessment All cities and countries are assumed to be equally vulnerable to terrorist attacks of a given magnitude (in terms of economic impact) in its first year. Similar to the vulnerability assessment for interstate conflict, the underlying theory is that there are few characteristics of a city that could ameliorate the impact of being attacked once it has happened. Cities that are more resilient recover more quickly however.

5.6 Consequence Analysis The initial impact to GDP due to terrorist attacks is determined from a combination of historical studies of economic impacts following attacks and subject matter expertise. Please see the CCRS study: A Risk Analysis Retrospective on the 2015 Paris Attacks at http://cambridgeriskframework.com/downloads.

The subsequent recovery from the initial GDP shock is determined by the socioeconomic resilience of each city. Characteristic recovery profiles for each resilience level were also determined from historical and scenario studies and from subject matter experts.

5.7 References

Global Terrorism Index 2017. (2017). Institute for Economics and Peace. URL: http://visionofhumanity.org/app/uploads/2017/11/Global-Terrorism-Index-2017.pdf

National Consortium for the Study of Terrorism and Responses to Terrorism (START). (2016). Global Terrorism Database. Retrieved from https://www.start.umd.edu/gtd Kelly, S.; Asante, S.; Jung, J. C. D.; Kesaite, V.; Woo, G.; A Risk Analysis Retrospective on the 2015 Paris Attacks; Working Paper 2016:1; Cambridge Risk Framework series; Centre for Risk Studies; University of Cambridge.

6 Social Unrest

6.1 Threat Description Human history is rife with instances of social unrest, insurgency and rebellion. To a great extent, civil order has always gone hand in hand with an element of civil disorder, and as society expands and develops, so too do the methods and arenas of protest and dissent. At its most extreme, social unrest can cause massive, widespread disruption and both its direct and indirect effects may be felt for a long period following. The decade-long French Revolution (1789-1799), for example, escalated from marketplace food riots to such a seismic upheaval of established ideologies and the socio-political regime that its impact on wider global history remains difficult to accurately quantify to this day. Historically, the relationship between socio-economic health and the risk of social unrest is a strong one. Wealth disparity, financial crisis and hyperinflation are regularly accompanied by stirrings of civil disorder; mass protest and resistance provide a significant route of public objection in the formation of economic policy and key political decisions. Cambridge Centre for Risk Studies has been studying the potential disruption to the economy, to social and business continuity, and for political upheaval and violence. Previous CCRS publications on social unrest risk include Millennial Uprising Social Unrest Stress Test Scenario at http://cambridgeriskframework.com/getdocument/22. Arab Spring Social unrest has gained prominence as a risk management issue for businesses with the surprising sequence of civil disorder in countries around the Middle East that became known as the Arab Spring movement. It began with protests about economic opportunity in late 2010 and within months had led to a change of leadership in countries ranging from Tunisia, Egypt, Libya and Yemen, with civil

Cambridge Centre for Risk Studies 26 Cambridge Global Risk Outlook 2018 Threat Profiles

disorder seen many other countries around the Middle East as ideas around potential change were communicated between the populations of neighbouring and similar countries. Armed uprisings and full civil war were one extreme end of the spectrum of social unrest. Internet communications played an important role. It was notable that young people were key participants. The outbreak caught a large number of political observers by surprise. Business interests were damaged by many of the outcomes, and there was considerable post-analysis about how to improve the prediction and warning of these socio-political events in the future. The major element of concern to businesses and to the global economy in general, was the ease and speed with which the movement spread from country to country. Prior to this, social unrest was a sporadic, occasional threat that could occur in one city or country, but this was a new phenomenon: a regional wave of change. Businesses started studying the globe for other potential regions where sudden waves of regime change might cascade through a region or set of countries with similar cultural values, and damage business interests in the same way. The Occupy movement Partly inspired by the Arab Spring uprisings, and during the slow economic recovery after the great financial crisis 0f 2008, there were a series of anti-austerity protests across Europe and United States in 2011 and 2012, which gained the name of the ‘Occupy’ movement, following a series of protest occupations of major city centres and iconic financial institutions (‘Occupy Wall Street’, ‘Occupy London’). The protest was about social and political inequality, using their slogan of “We are the 99%” to highlight statistics of the richest 1% of the population owning a large proportion of the total wealth in a period of polarizing disparity. Again, many of the participants and activists were disenfranchised young adults. In Spain and Portugal the movement was known as the indignados, and had its origins in movements to protest high unemployment, such as Juventud Sin Futuro (Youth Without a Future). The Occupy movement succeeded in coordinating political rallies across many cities and countries. The movement used social media such as Facebook, Twitter, n-1 and Google Drive to great effect to coordinate events world-wide. On 15 October, 2011, Occupy arranged a global protest in which millions of people took part, taking to the streets of 950 cities in a single day. Social unrest emerging as a systemic risk What is different and new about the episodes of civil disorder in the early 21st century is their systemic nature: The fact that multiple countries simultaneously expressed dissatisfaction and sought change. The protest movements occurred in many places at once, amplifying the disruption caused. This coordination is enabled by uncontrollable social media and new democratized communication – it is now possible for ideas and actions to be spread through the twittersphere and cell phone messaging to bring thousands of people together in coordinated protests. Social unrest is now a systemic threat, capable of destabilizing many countries at once, posing potential threats to entire regions of the world, or demographic segments of the population. This is changing the way that businesses are thinking about their risk of political instability. Youth unemployment as a key driver Since the 2007 economic crisis and the advent of the ‘Great Recession’, record-high rates of unemployment, eviction and income inequality have led to waves of public demonstration and strike action throughout the western world. Exasperated by modern technology, mobile news and new social medias, local social unrest may spread worldwide in a matter of hours and is no longer confined to streets and sit-ins. Large crowds can now gather in the virtual sphere, participating in online protests and carrying out acts of computerised sabotage. Social unrest now exists in a myriad of environments and may appeal to the entirety of the internet, enlisting people across the globe as its intended audience. Types of social unrest Social unrest encompasses a broad spectrum of public dissent, ranging from peaceful protest to armed insurrection. Within this broad understanding, stages of escalation can be categorised as follows, and modelled in the Cambridge Global Risk Index as escalating severities of Local Impact Severity. Level 1: Social Unrest

Cambridge Centre for Risk Studies 27 Cambridge Global Risk Outlook 2018 Threat Profiles

Social Unrest describes activities of dissent by sectors of the population to challenge the established authorities. Peaceful methods of protest such as anti-establishment commentary, media campaigns, and petitions often accompany marches, protests, disobedience and non-violent resistance. Level 2: Civil Disorder Social unrest can escalate into civil disorder, a law enforcement term to describe activities by a group of people that cross the boundary of illegality (e.g. causing disturbances and damaging property). Groups in this stage lack formal leadership and/or explicit aims or objectives, and violence tends to be spontaneous and short-lived with improvised weapons. The escalation out of civil disorder to a more concentrated or organized violent movement is referred to as mob rule, during which law enforcement cannot be maintained and authorities are unable to restore control over a particular geographical area, or for a particular period of time. Level 3: Civil War When a system of leadership starts to emerge within dissenting groups and movements begin to formulate specific aims and objectives, an all-out rebellion can ensue. This is the most extreme form of social unrest and involves uprising and/or insurrection with a view to overthrowing the ruling regime. Rebellion that is resisted by the authorities develops into civil war. Level 1 of Social Unrest is, compared to Levels 2 and 3, a more democratic expression of discontent and decidedly more peaceful. While these expressions of unrest are related, the nature of each is different. Countries and cities which experience high levels of Level 1 Social Unrest may not experience Level 2 and Level 3 violent expressions of Civil Conflict. One example is France which has a strong history of labour protests but less violent disorder (in modern times). However, high levels of both violent and non-violent unrest are seen in other countries such as Turkey which include demonstrations in Taksim Square in 2013 to a coup d’etat in 2015. Due to this differentiation, Level 2 and 3 of Social Unrest are categorized as Civil Conflict in a separate threat category.

6.2 Mapping the Threat Categorization of each country by its threat of social unrest is derived from the Economist Intelligence Unit.27 A subject matter expert graded certain countries higher or lower given changes to the current landscape of social unrest from when the EIU analysis was conducted in 2014, as well as from knowledge of social unrest indicators not explicitly modelled by the EIU index.

The EIU index is a derivation of the Political Instability Index created in 2009.28 It is based on factors such as a democratic or authoritarian regime type, income inequality, quality of government, trust in government, economic distress and a history of unrest.

Threat Assessment Grade No. of Examples Countries A Very High Risk 15 Afghanistan, Turkey, Zimbabwe B High Risk 30 Brazil, Cambodia, Spain C Moderate Risk 39 Ireland, South Korea, United States D Low Risk 18 Australia, Canada, Sweden E Very Low Risk 5 Austria, Norway

6.3 Local Impact Severity Description Each city is analysed for the GDP impact and likelihood of experiencing the following characteristic social unrest scenarios.

27 The Economist Intelligence Unit: Social Unrest in 2o14. (2014) URL: https://www.economist.com/blogs/theworldin2014/2013/12/social-unrest-2014 28 The Economist Intelligence Unit: Political Instability Index: Vulnerability to social and political unrest. (2009). URL: http://viewswire.eiu.com/index.asp?layout=VWArticleVW3&article_id=874361472

Cambridge Centre for Risk Studies 28 Cambridge Global Risk Outlook 2018 Threat Profiles

LIS Description SU1 Civil Unrest causes riots and protests in the streets for months; violent confrontations with police

6.4 Quantifying the Threat The annual likelihoods for the characteristic threat assessment grades were derived largely from historical frequency analysis of protests and riots by the Cambridge Centre for Risk studies.

Threat Assessment Grade Annual Likelihood Return Period A Very High Risk 0.07174 14 B High Risk 0.01433 70 C Moderate Risk 0.00267 375 D Low Risk 0.00068 1462 E Very Low Risk 0.00026 3802

6.5 Vulnerability Assessment All cities and countries are assumed to be equally vulnerable to, i.e. experience the same initial impact to GDP, from social unrest. There may be scope in the future to further develop this threat model and distinguish city-level vulnerability.

6.6 Consequence Analysis The initial impact to GDP due to social unrest is determined from a combination of historical studies of economic impacts following attacks and subject matter expertise. For more detail on an economic scenario analysis of social unrest, please see: Millennial Uprising Social Unrest Stress Test Scenario at http://cambridgeriskframework.com/downloads.

The subsequent recovery from the initial GDP shock is determined by the socioeconomic resilience of each city. Characteristic recovery profiles for each resilience level were also determined from historical and scenario studies and subject matter experts.

6.7 References Bowman, G.; Caccioli, F.; Coburn, A.W.; Hartley, R.; Kelly, S.; Ralph, D.; Ruffle, S.J.; Wallace, J.; (2014). Stress Test Scenario: Millennial Uprising Social Unrest Scenario; Cambridge Risk Framework series; Centre for Risk Studies, University of Cambridge.

The Economist Intelligence Unit: Social Unrest in 2o14. (2014). URL: https://www.economist.com/blogs/theworldin2014/2013/12/social-unrest-2014

The Economist Intelligence Unit: Political Instability Index: Vulnerability to social and political unrest. (2009) URL: http://viewswire.eiu.com/index.asp?layout=VWArticleVW3&article_id=874361472

7 Civil Conflict

7.1 Threat Description Civil conflicts have the potential to cause large scale humanitarian disasters as well as catastrophic economic shocks. Large refugee flows and destruction of infrastructure are some of the devastating impacts seen in on-going sub-national conflicts, notably now in Syria, Yemen and Myanmar. As described in the Social Unrest section, the Centre for Risk Studies considers Civil Conflicts as escalations of a Level 1 Social Unrest into large-scale events of a violent nature.

The differences between terrorism and civil conflict often lead to contentious debate. The Centre for Risk Studies has taken the accepted view that terrorism is an act of violence usually perpetrated

Cambridge Centre for Risk Studies 29 Cambridge Global Risk Outlook 2018 Threat Profiles

against non-combatants to achieve a political aim. Terrorism can and often seeks to influence policy, and does not have to be indigenous nor challenge the direct authority of a non-local government. On the other hand, civil conflict generally consists of well organised groups who use protracted violence and conflict to challenge the established authority of a government or state. These groups are mainly concerned with localised issues, and although terrorism may be utilised as a tactic, clearly defined issues guide their conduct.

Civil conflicts can involve state actors and non-state actors, but rule out interstate conflicts that cross domestic boundaries. Examples include violence between cartels, paramilitary groups, left-wing militants and gangs in Colombia, El Salvador and Mexico over drugs, land and natural resources. Conflicts may also manifest as a secessionist movement. While separatist struggles are similarly underscored by claims to land and resources, they also exhibit ideological differences. Secession and autonomy conflicts in recent history include the Kurdistan Workers’ Party (PKK) movement in Turkey and Baloch nationalists’ movement in Pakistan.

Source: https://www.hiik.de/en/konfliktbarometer/pdf/ConflictBarometer_2016.pdf

7.2 Mapping the Threat The country level risk assessment of Civil Conflict is based off the Global Conflict Risk Index’s (GCRI) Probability of Violent Internal Conflict from the Index for Risk Management (INFORM). GCRI is a conflict index produced by the European Commission Joint Research Centre; its output a part of the hazard assessment of the Index for Risk Management (INFORM), a risk index created by the Inter- Agency Standing Committee and the European Commission.29 The GCRI Probability of Violent Internal Conflict produces a risk score which measures the relative likelihood of a country to incur a national and/or sub-national conflict, based on political, security, social, economic and geographical factors and also from historical data on outbreaks of violence.

Risk Area Concept Indicator Source

Political Regime type Regime Type CSP

Lack of Democracy CSP

Government Effectiveness World Bank

29 INFORM Index for Risk Management http://www.inform-index.org/ Note that parts of the INFORM index were also used for the resiliency assessment of the Cambridge Global Risk Index.

Cambridge Centre for Risk Studies 30 Cambridge Global Risk Outlook 2018 Threat Profiles

Regime Level of Repression CIRI performance Empowerment Rights PTS Security Current conflict Recent Internal Conflict HIIK; UCDP/PRIO situation Neighbouring with HVC HIIK; UCDP/PRIO History of conflict Years since HVC HIIK; UCDP/PRIO

Social Social cohesion Corruption World Bank and diversity Ethnic Power Change ETH Zurich

Ethnic compilation ETH Zurich

Transnational Ethnic Bonds CIDCM

Public security Homicide Rate UNODC and health Infant Mortality UNICEF

Economy Development and GDP per capita World Bank distribution Income inequality World Bank

Openness World Bank

Provisions and Food Security FAO Employment Unemployment ILOSTAT

Geography Geographic Water Stress WRI and challenge Oil Production World Bank Environment

Structural Constraints BTI

Demographics Population Size World Bank

Youth Bulge UNDESA Source: Global Conflict Risk Index. http://conflictrisk.jrc.ec.europa.eu/Methodology

Threat Assessment No. of Examples Grade Countries

A Very High Risk 8 Afghanistan, Yemen B High Risk 13 Bangladesh, Egypt C Moderate Risk 37 Colombia, Mexico, Russia D Low Risk 16 Argentina, United Kingdom E Very Low Risk 32 Australia, Iceland

7.3 Local Impact Severity Definitions Each city is analysed for the GDP impact and likelihood of experiencing the following characteristic civil conflict scenarios:

LIS Description CC1 See Social Unrest CC2 Incidents of sectarian fighting between armed gangs and private militias in the streets of the city for multiple years CC3 Violence involves months of street fighting between well-organized and well-equipped armies using heavy weaponry.

7.4 Quantifying the Threat The likelihood assessment of each level of civil conflict was approximated using historical frequency of violent internal conflicts such as civil wars. Primary data sources include the Heidelberg Institute for

Cambridge Centre for Risk Studies 31 Cambridge Global Risk Outlook 2018 Threat Profiles

International Conflict Research30 and Uppsala University Depart of Peace and Conflict Research Conflict Data Program31. Both data sources are used in the Global Conflict Risk Index. All data is analysed by a subject matter specialist to determine future likelihoods compared against the long term baseline.

Annual Likelihood Return Period Threat Assessment Grade CC2 CC3 CC2 CC3 A Very High Risk 0.04337 0.02169 23 46 B High Risk 0.01752 0.00584 57 171 C Moderate Risk 0.00876 0.00292 114 343

D Low Risk 0.00088 0 1142

E Very Low Risk 0 0

7.5 Vulnerability Assessment The economic vulnerability of countries and cities to civil conflict is assumed to be related to its physical vulnerability, i.e. the quality of buildings and infrastructure. Each country is assigned to a vulnerability level based off the earthquake vulnerability analysis (see chapter 8). The initial shocks for each vulnerability level is specific to the civil conflict consequence analysis.

Vulnerability No. of Examples Countries 1 Very Strong 4 Chile, New Zealand, Japan, United States 2 Strong 24 Australia, Canada, Germany, Portugal 3 Moderate 35 Argentina, Bulgaria, Tunisia, Venezuela 4 Weak 28 Bangladesh, Libya, Thailand 5 Very Weak 16 Angola, Sudan, Zimbabwe

7.6 Consequence Analysis The initial impact to GDP due to civil conflict is determined from a combination of in-depth scenario and historical studies and subject matter expertise. The impacts of civil conflict are an often violent extension of social unrest, as described in the Millennial Uprising Social Unrest Stress Test Scenario report at http://cambridgeriskframework.com/downloads.

The subsequent recovery from the initial GDP shock is determined by the socioeconomic resilience of each city. Characteristic recovery profiles for each resilience level were also determined from historical and scenario studies as well as from subject matter experts.

7.7 References Heidelberg Institute for International Conflict Research (2017): Conflict Barometer 2016, Heidelberg. URL: https://www.hiik.de/en/konfliktbarometer/pdf/ConflictBarometer_2016.pdf

Halkia, S., Ferri, S., Joubert-Boitat, I., Saporiti, F. (2017). Conflict Risk Indicators: Significance and Data Management in the GCRI, EUR 28860 EN, Publications Office of the European Union, Luxembourg. URL: http://publications.jrc.ec.europa.eu/repository/handle/JRC107996

Uppsala Conflict Data Program (2017) UCDP Conflict Encyclopedia: www.ucdp.uu.se, Uppsala University

30 Heidelberg Institute for International Conflict Research Conflict Database (2017) URL: https://hiik.de/data-and- maps/conflict-database/ 31 Uppsala Conflict Data Program (2017) UCDP Conflict Encyclopedia: www.ucdp.uu.se, Uppsala University

Cambridge Centre for Risk Studies 32 Cambridge Global Risk Outlook 2018 Threat Profiles

Part C Natural Catastrophe and Climate

8 Earthquake

8.1 Threat Description Earthquake risk is highly localized and depends on the seismic fault structures in and around the city. This study does not attempt a detailed seismic source analysis but uses United States Geological Survey assessments of the design load spectral response acceleration at the centroid of each city, together with peak ground acceleration mappings from United Nations Environmental Program, and a historical catalogue of earthquake events from the Significant Earthquake Database of US National Geophysical Data Center, to categorize cities by seismic hazard.

8.2 Mapping the Threat Earthquake hazard for the cities is determined from the United States Geological Survey (USGS) seismic design criteria for the location of each city, represented by the latitude and longitude coordinates of its centroid.

The USGS design tool32 provides the strength of earthquake shaking that would be expected for that location. Specifically, it provides the severity parameters for seismic design of structures using the International Building Code (§1613) and similar standards (e.g., the ASCE/SEI 7 Standard and the U.S. Department of Defense Unified Facilities Criteria). It provides SS and S1 (spectral response acceleration at 0.2 and 1.0 seconds, respectively, for 5% of critical damping) values derived using seismic hazard data from numerous sources. All values provided are for Site Class B.

S1 is an almost exact multiple of SS for almost all cities, so SS was used as the earthquake threat assessment metric. SS essentially applies to low-rise structures (around 2 storeys) and S1 typically applies to medium rise buildings (around 10 storeys). SS is a more representative value of the majority of the property under threat.

Cities were assigned into Earthquake Threat Assessment Groups (TAG). Long-period effects (destructive effects on medium or high risk buildings that are caused from distant earthquakes resonating through deep soft soil structures) are known effects for certain cities, like Mexico City and Bucharest. These can be seen in high S1 values relative to the SS assignments, and these are adjusted for by upgrading these cities into a higher TAG category.

Threat USGS SS No. of Examples Assessment Grade Range Cities A (2-4) 12 Tehran, Wellington B (1-2) 46 Bogota, Istanbul, Tokyo C (0.5-1) 54 Bratislava, Lahore D (0.25-0.5) 47 Budapest, Kuala Lumpur, Xiamen E (0.1-0.25) 70 Guangzhou, Rotterdam, Seoul F (<0.1) 50 Buenos Aires, Dallas, Moscow

8.3 Local Impact Severity Definitions Earthquake magnitude scales measure the size of the event by the energy release. Surface Wave Magnitude (Ms) is a commonly accepted magnitude scale that best correlates with destructive power of an earthquake. Earthquake catalogues have good data on the magnitudes of past events, so recurrence rates are best assessed in terms of earthquake magnitude. The magnitudes of earthquakes that are of most concern are ‘Large Magnitude Earthquakes’ (Ms 6.0-7.0) and ‘Great Earthquakes’ (Ms

32 https://earthquake.usgs.gov/hazards/designmaps/usdesign.php

Cambridge Centre for Risk Studies 33 Cambridge Global Risk Outlook 2018 Threat Profiles

7.0-8.0). Earthquakes as large as Ms 9.0 can occur – mainly offshore in deep subduction zones. To assess the risk of cities from earthquake effects, we consider the impact of Large Magnitude Earthquakes and ‘Great Earthquakes’ occurring close to cities.

The impact of an earthquake, in terms of how much destruction would be caused to a city, is analysed by considering the footprint of the intensity of ground shaking produced by an earthquake of a specific magnitude. Destructive shaking can be measured in a variety of metrics, from Peak Ground Acceleration to Spectral Acceleration and more qualitatively by Intensity scales. Intensity scales describe approximate destruction effects but are useful because their historical coverage is much more comprehensive than any other metric. There are many variants of the original Modified Mercalli Intensity (MMI) scale (including the MSK, EMS, IMS) but for simplicity we refer here to them generically as ‘MMI’. MMI intensities are typically described with Roman numerals, and levels of VII (7) to X (10) are generally the destructive intensity levels of concern.

We consider the geographical footprint of an earthquake and its proximity to the city, to define three characteristic scenarios of concern for earthquake-prone cities. The possible ways in which the above Local Impact Severity scenarios can happen are defined below. As each defined earthquake magnitude can occur in multiple ways, e.g. centroid of city experiences VII, the combinations of possible earthquakes are necessary to define the likelihood of the events as described.

LIS Description EQ1 A 'Large Magnitude Earthquake' (Ms6.5) within the city boundaries. Centroid of city experiences VII (PGA 250-400) EQ2 A 'Great Earthquake' (Ms7.0) with its epicentre close to the edge of the city, just outside its boundaries. Centroid of city experiences VIII (PGA 400-600) EQ3 A 'Great Earthquake' (Ms7.5) occurring at shallow depth with its epicentre close to the centre of the city. Centroid of city experiences IX (PGA 600-1000)

8.4 Quantifying the Threat We define the probability of occurrence of these scenarios at each city, according to the threat assessment grade classification of the city. This is a simplification of hazard assessment and is not intended as a substitute for a detailed analysis of seismic hazard for each city.

The historical frequency of occurrence of earthquakes of different magnitudes is assessed for cities in each of the bands of the TAG, for a region (a buffer zone) of 250km radius around the centroid of each city. This treats the region around each city as a source zone of uniform seismicity, in which earthquakes are assumed equally likely to occur at any location (i.e. we do not consider individual fault locations or detailed processes of seismic sources around the city). The buffer zone is large enough to account for uncertainties in epicentral location for events with large magnitudes capable of affecting the cities with damaging intensities. The zones around all the cities of each TAG level are aggregated and the catalogue of historical earthquakes observed for the past 50 years (since 1964) to provide a sufficient statistical large observation to derive stable magnitude frequency recurrence relationships.

Figure 8.1 shows the magnitude frequency recurrence within a 100km radius of the city that are derived from this study for a city in each of the Threat Assessment Grades A-C.

Cambridge Centre for Risk Studies 34 Cambridge Global Risk Outlook 2018 Threat Profiles

0.025

0.020

0.015 A B 0.010 C

0.005

0.000 Ms>6.0 Ms>6.5 Ms>7.0 Ms>7.5

Figure 8.1 Earthquake magnitude frequency recurrence within a 100km radius of a city in each TAG Source: CCRS analysis, National Geophysical Data Center / World Data Service (NGDC/WDS): Significant Earthquake Database. National Geophysical Data Center, NOAA. doi:10.7289/V5TD9V7K

A simple magnitude intensity attenuation relationship33 was used to assume the distance away from the city that would cause the effects of each characteristic scenario. This is summarized in table 8.1.

Distance from city (km) to cause Characteristic Scenario

Earthquake of Magnitude Characteristic Scenario Ms=6.0-6.5 Ms=6.5-7.0 Ms=7.0-7.5 Ms>7.5 EQ1 50 70 100 150 EQ2 10 30 60 90 EQ3 0 0 30 75

Table 8.1 Distance away from the city for earthquakes of different magnitudes to cause characteristic scenarios The probability of occurrence of earthquakes of different magnitudes within the seismic source zones, within the various radii of the cities is calculated to provide the probabilities of experiencing the three local impact severities. The annual likelihood is estimated from the historical analysis of earthquake recurrence within a 100km of each city from figure 8.1. Each city is assigned an annual probability of experiencing the three Characteristic Scenarios, according to their Threat Assessment Grade. This is a simplified model of earthquake hazard analysis, but is appropriate for assessing a first-order assessment of economic impact of threat to be used in a comparative analysis of GDP@Risk between cities and across different threats. This simplified model could be improved and a more refined hazard modelling substituted in due course.

Annual Likelihood Return Period Threat Assessment USGS SS EQ1 EQ2 EQ3 EQ1 EQ2 EQ3 Grade Range A (2-4) 0.0124 0.0075 0.0031 80 134 321 B (1-2) 0.0066 0.0042 0.0018 152 241 562

33 Average intensity attenuation relationships from analysis of 53 shallow-depth onland earthquakes, in Coburn & Spence (2002) p247.

Cambridge Centre for Risk Studies 35 Cambridge Global Risk Outlook 2018 Threat Profiles

C (0.5-1) 0.0021 0.0012 0.0002 483 864 5333

D (0.25-0.5) 0.0001 0 0 12000

E (0.1-0.25) 0 0 0

F (<0.1) 0 0 0

8.5 Vulnerability Assessment The economic impact model considers the physical vulnerability of the city, in terms of the quality of the building stock and the strength to withstand earthquake forces. Cities of high vulnerability (predominantly weak building stock) suffer high levels of damage in an earthquake. The damage calculation incorporates vulnerability estimation for the cities in earthquake areas.

Cities are classified into five grades of physical vulnerability using categorizations of earthquake vulnerability for 47 countries defined for the International Macroseismic Scale34. Countries like Chile, New Zealand, USA and Japan that have strong and well-enforced seismic building codes are rated as ‘1 very strong’ with countries that have informal building sectors and low resistance to earthquakes, such as Iran, Pakistan and Algeria, rated as ‘4 weak’.

As an input to the vulnerability assessment and GDP loss estimation, damage ratios are estimated for cities with different mixes of buildings with different vulnerability levels. Building types range from artisan or vernacular construction (type A) through to engineered structures to high levels of seismic code (type DC). Statistical distributions of the damage observed for these different building types at different intensity levels of earthquake ground shaking, derived from a large number of post- earthquake damage surveys, are used as inputs35. A process of estimating the physical vulnerability of a city, in terms of its mix of different qualities of construction type, on the overall levels of destruction, provides a mean damage ratio (average repair and replacement cost of the damage suffered).

Vulnerability (Mean damage ratios at MMI Intensity) Intensity Type A: (V Weak) Type B Type DC (V Strong) 10 (X) 0.82 0.53 0.23 9 (IX) 0.72 0.43 0.14 8 (VIII) 0.45 0.25 0.04 7 (VII) 0.21 0.11 0.01

Table 8.2 Typical differences in damage (mean damage ratios) for building types of different vulnerability at earthquake intensities Source: CCRS Analysis The analysis of the average mean damage ratio for each city forms the basis for categorizing the cities and countries into physical vulnerability grades. These assessments were also used as a basis for vulnerability to other physical threats as the building types and mix are vulnerable to other threat impacts. Physical Vulnerability No. of Countries Examples 1 Very Strong 4 Chile, New Zealand, Japan, United States 2 Strong 24 Australia, Canada, Germany, Portugal 3 Moderate 35 Argentina, Bulgaria, Tunisia, Venezuela 4 Weak 28 Bangladesh, Libya, Thailand 5 Very Weak 16 Angola, Sudan, Zimbabwe

34 Foulser-Piggott & Spence (2013) 35 Damage distributions for different building types in Coburn & Spence (2002), pp323-333.

Cambridge Centre for Risk Studies 36 Cambridge Global Risk Outlook 2018 Threat Profiles

8.6 Consequence analysis An earthquake hitting a city would have significant consequences for the economic output of the city in a number of ways. The destruction of the city is a key component, in damaging the physical means of production (factories and commercial property, equipment, infrastructure) and disrupting the social structure, making workers homeless, destroying transportation networks, and affecting the morale of the population. Most of the economic impact is closely related to the physical damage caused by the earthquake. The analysis considers the impact of the earthquake on the amount of physical destruction that it would cause to the building stock of the city, and then translates that into the economic impact on the city’s output.

To do this, we consider an idealized city and how the footprint of the earthquake might affect it. The idealized city consists of around 2,500 square kilometres of residential suburbs (approximately 50 km x 50 km or about 28 km radius), with a commercial area of around 400 square kilometres (20 km x 20 km or about 11 km radius) and a central business district of around 10 square kilometres (3 km x 3 km or about 2 km radius). Cities vary in their areas and geometry considerably, but this conforms to a moderate sized urban structure. We have used a relative ratio of the property value per unit area of 1:10:100 for residential:commercial:CBD.

Each characteristic earthquake local impact severity scenario can be described as an earthquake footprint of concentric ellipses of various sizes, determined by average intensity attenuation distances for earthquakes of representative magnitudes. Overlaying these footprints on the idealized city provides an estimate of the levels of destruction that would result. The mean damage ratios as described in section 8.5 provide estimates for the amount of economic damage suffered from the earthquake footprints.

8.7 References Coburn, A. and Spence, R. (2002) Earthquake Protection, Second Edition, John Wiley & Sons, Ltd, Chichester, UK. doi: 10.1002/0470855185.fmatter

National Geophysical Data Center / World Data Service (NGDC/WDS): Significant Earthquake Database. National Geophysical Data Center, NOAA. doi:10.7289/V5TD9V7K

Carreño, M L, Cardona, O.D. and Barbat, A.H. (2012). “New methodology for urban seismic risk assessment from a holistic perspective” Bulletin of Earthquake Engineering. 40: 137-172. Foulser-Piggott, R.; Spence, R.; 2013; ‘Extending EMS-98 for more convenient application outside Europe II: Development of the International Macroseismic Scale’; Paper No. 382 in Vienna Congress on Recent Advances in Earthquake Engineering and Structural Dynamics 2013 (VEESD 2013).

9 Tropical Windstorm and Temperate Windstorm

9.1 Threat Description Windstorm threats consist of tropical storms and temperate windstorms. By far the most destructive storms are tropical storms, known as hurricanes, cyclones, and typhoons in different parts of the world. Hurricanes, typhoons and cyclones are all tropical windstorms, characterized by low pressure spiralling wind systems capable of generating strong wind speeds, very intense rainfall, and major sea water surges into the coast. These storms can be very destructive and damaging to the economy. Examples of devastating windstorms include Hurricane Katrina in the US, 2005, Typhoon Mireille in Japan in 1991, and European Windstorm Lothar in 1991. Temperate windstorms are windstorms outside the tropical region, although the storm could have originated as a tropical windstorm.

The mapping here shows the Pacific Research Center zoning for the likelihood of hurricane force wind speeds from tropical storms with a 10% probability of occurring within the next 10 years.

Cambridge Centre for Risk Studies 37 Cambridge Global Risk Outlook 2018 Threat Profiles

Source: United Nations Office for the Coordination of Humanitarian Affairs - Regional Office for Asia and the Pacific (OCHA ROAP), derived from the Munich Reinsurance Company's World Map of Natural hazards. URL: http://www.preventionweb.net/files/3856_ocharoapstormsv5110501.pdf

There is also the possibility of multiple storms occurring within a region during a particular storm season. Tropical windstorms occur over warm oceans, in the tropics, during a specific period of the summer, when sea surface temperatures are high enough to generate the storm system energy.

9.2 Mapping the Threat We assessed the windstorm risk of each city using data from the Global Risk Data Platform Tropical cyclone frequency of Saffir-Simpson category 5 1970-2009 from UNEP/DEWA/GRID-Europe.36 This dataset contains the geographical distribution of peak velocity gusts (km/h) and maximum Saffir- Simpson category cyclones for return periods: 50, 100, 250, 500, and 1000 years. The combination of these two metrics was used to assign threat assessment grades to each city. Stronger windspeeds (A to D) are classifications for tropical windstorm while lower windspeeds (E-H) are classifications for temperate windstorm.

250yr RP 250yr RP No. of Threat Assessment Grade peak gusts max Saffir- Examples Cities (km/h) Simpson A Very High Threat of 290-350 5 6 Hiroshima, Osaka Hurricane-Strength Winds B High Threat of Hurricane- 250-290 5 12 Daegu, Gwangju, Strength Winds Shanghai C Moderately High Threat of 215-250 4-5 15 Calcutta, Hong Kong, Hurricane-Strength Winds Seoul D Moderate Threat of 190-215 3-4 9 Houston, Nanjing Hurricane-Strength Winds

36 UNEP/DEWA/GRID-Europe. Tropical cyclone frequency of Saffir-Simpson category 5 1970-2009. http://preview.grid.unep.ch/index.php?preview=data&events=cyclones&evcat=2&lang=eng

Cambridge Centre for Risk Studies 38 Cambridge Global Risk Outlook 2018 Threat Profiles

E Threat of Storm-Force 125-190 1-2 33 Atlanta, Dhaka, Winds Wuhan F Threat of Gale-Force Winds 70-125 0-1 38 Bangkok, Detroit, Panama City G Rare Threat of Strong 0-70 0 42 Abu Dhabi, Winds Johannesburg, Seattle H Low Threat of Strong 124 Ankara, Frankfurt, Winds Quito

9.3 Local Impact Severity Definitions Each city is analysed for the GDP impact and likelihood of experiencing the following characteristic windstorm scenarios:

LIS Description HU1 Category 1 Hurricane, windspeed 118-153 km/hr HU2 Category 3 Hurricane, windspeed 178-209 km/hr HU3 Category 5 Hurricane, windspeed >250 km/hr

LIS Description TW1 Wind Storm of Beaufort Scale 10: Storm or Whole Gale (windspeeds of over 89 km/hr) TW2 Wind Storm of Beaufort Scale 11: Violent Storm (windspeeds of over 103 km/hr) TW3 Wind Storm of Beaufort Scale 12: Hurricane (windspeeds of over 118 km/hr)

9.4 Quantifying the Threat The likelihood of windstorms for each characteristic scenario affected thing cities in each TAG was based off historical analysis of windstorms. Major sources used include UNEP/DEWA/GRID-Europe cyclone frequency data and EM-DAT, the International disaster database from the Centre for Research on the Epidemiology of Disasters.37 EM-DAT data contains information on previous storm events, including: occurrences since 1900, deaths, number of people affected, number of people injured, amount homeless and total damage ('000$).

Annual Likelihood Return Period Threat Assessment Grade HU1 HU2 HU3 HU1 HU2 HU3 A Very High Threat of Hurricane- 0.2175 0.0464 0.00725 5 22 138 Strength Winds B High Threat of Hurricane- 0.087 0.02175 0.006525 11 46 153 Strength Winds C Moderately High Threat of 0.0435 0.0116 0.00174 23 86 575 Hurricane-Strength Winds D Moderate Threat of Hurricane- 0.0174 0.00435 0.000725 57 230 1379 Strength Winds

Annual Likelihood Return Period Threat Assessment Grade TW1 TW2 TW3 TW1 TW2 TW3 E Threat of Storm-Force Winds 0.00512 0.00202 0.00040 195 495 2495 F Threat of Gale-Force Winds 0.00338 0.00134 0.00027 295 745 3745 G Rare Threat of Strong Winds 0.00224 0.00089 0.00018 445 1120 5620 H Low Threat of Strong Winds 0 0 0

37 EM-DAT. http://www.emdat.be/database

Cambridge Centre for Risk Studies 39 Cambridge Global Risk Outlook 2018 Threat Profiles

9.5 Vulnerability Assessment The economic vulnerability of countries and cities to tropical windstorms is assumed to be related to its physical vulnerability, i.e. the quality of buildings and infrastructure. Each city is assigned to a vulnerability level based off the earthquake vulnerability analysis (see Section 8), with several adjustments made based on subject matter specialist knowledge on windstorm damage.

While the assigned vulnerability category is based off the earthquake analysis, the initial shocks for each vulnerability level is specific to the windstorm consequence analysis.

Vulnerability Level No. of Cities Examples 1 Very Strong 42 Boston, Shanghai, Wellington 2 Strong 58 Amsterdam, Hamburg, Shenzhen 3 Moderate 100 Daegu, Manchester, Zagreb 4 Weak 63 Almaty, Esfahan, Pretoria 5 Very Weak 16 Accra, Harare, Port-Au-Prince

Relative to tropical windstorm, temperate windstorm is a minor peril. At this stage of model development, cities are currently not differentiated for vulnerability to temperate windstorms. All cities are assumed to receive the same impact to GDP following an event of a given size.

9.6 Consequence Analysis The tropical windstorm consequence analysis includes damage from flooding and surges while for temperate windstorm, the primary hazard is the high wind speed. The initial impact to GDP due to windstorm for each vulnerability level is determined through a combination of in-depth scenario modelling, historical studies and subject matter expertise. These studies were also used to determine the subsequent recovery from the initial GDP shock for cities within each socioeconomic resilience level.

This analysis is supported by subject matter specialists in conjunction with data and case studies that exist on economic damage from natural disasters, including from EM-DAT, a comprehensive disaster database by the Centre for Research on the Epidemiology of Disasters. Other validation methods include relativity analysis in comparison to economic damage from other threats.

9.7 References United Nations Office for the Coordination of Humanitarian Affairs - Regional Office for Asia and the Pacific (OCHA ROAP), derived from the Munich Reinsurance Company's World Map of Natural hazards. URL: http://www.preventionweb.net/files/3856_ocharoapstormsv5110501.pdf

D. Guha-Sapir, R. Below, Ph. Hoyois - EM-DAT: The CRED/OFDA International Disaster Database – www.emdat.be – Université Catholique de Louvain – Brussels – Belgium.

Peduzzi, Pascal. UNEP/DEWA/GRID-Europe. Tropical cyclones surges frequency 1975-2007. Geneva, Switzerland. URL: http://preview.grid.unep.ch/index.php?preview=data&events=surges&evcat=2&lang=eng

10 Tsunami

10.1 Threat Description Coastal cities have a threat of tsunami risk when a major earthquake, submarine landslide, or oceanic meteorite triggers sea waves that wash ashore. The most severe tsunami threat comes from large magnitude earthquakes that occur a short distance off-shore and that trigger large run-up waves that may over-top city flood defenses. Historical records and tsunami hazard assessments such as those from the US National Oceanic and Atmospheric Administration indicate the most tsunami-prone

Cambridge Centre for Risk Studies 40 Cambridge Global Risk Outlook 2018 Threat Profiles

coastlines, such as those mapped here. Tsunami threat for each city is analysed by GDP impact and likelihood of experiencing characteristic scenarios. A tsunami or seismic sea wave, is an ocean wave generated by ocean bed disturbances far from the coast, breaking on the shore. The wave can be very large – it builds in height as it runs ashore – and is very destructive. Waves of several tens of metres in height have been recorded. Earthquakes, volcanic eruptions and other underwater explosions (including detonations of underwater nuclear devices), landslides, glacier calvings, meteorite impacts and other disturbances above or below water all have the potential to generate a tsunami.

10.2 Mapping the Threat Maps of historical tsunami events from The Tsunami Laboratory of the Institute of Computational Mathematics and Mathematical Geophysics38 and National Oceanic and Atmospheric Administration NOAA catalogue of tsunami events 39 were used to assess the historical risk of tsunami to each city. The threat assessment grades, shown below, were derived from these map layers and the coastal distance of each city.

Threat Assessment Grade No. of Cities Examples A Major Tsunami Threat from Offshore 9 Jakarta, Osaka, Tokyo Great Earthquakes B Tsunami Threat from Offshore Large 5 Hiroshima, Los Angeles Magnitude Earthquakes C Rare Tsunami Threat from Distant 2 Hong Kong Earthquakes D Minimal or No Tsunami Threat 263 Belgrade, Yerevan

10.3 Local Impact Severity Definitions Each city is analysed for the GDP impact and likelihood of experiencing the following characteristic tsunami scenarios:

LIS Description TS1 Tsunami with 3m run-up TS2 Tsunami with 6 m run-up TS3 Tsunami with 12 m run-up

10.4 Quantifying the Threat The annual likelihood estimates of each characteristic tsunami event and threat assessment grade are based off historical frequency analysis of tsunami events. Databases used include the disaster database EM-DAT.

Annual Likelihood Return Period Threat Assessment Grade TS1 TS2 TS3 TS1 TS2 TS3 A Major Tsunami Threat from 0.010 0.002 0.001 100 500 1000 Offshore Great Earthquakes B Tsunami Threat from Offshore Large 0.005 0.001 0.000 200 1000 10000 Magnitude Earthquakes C Rare Tsunami Threat from Distant 0.002 0.000 0.000 500 Earthquakes D Minimal or No Tsunami Threat 0.000 0.000 0.000

38 The Tsunami Laboratory of the Institute of Computational Mathematics and Mathematical Geophysics URL: http://tsun.sscc.ru/tgi_1.htm 39 National Oceanic and Atmospheric Administration JetStream Tsunami Locations & Occurrences. URL: http://www.weather.gov/jetstream/locations

Cambridge Centre for Risk Studies 41 Cambridge Global Risk Outlook 2018 Threat Profiles

10.5 Vulnerability Assessment At this stage of model development, cities are currently not differentiated for vulnerability to temperate windstorms. All cities are assumed to receive the same impact to GDP following an event of a given size.

10.6 Consequence Analysis The initial impact to GDP due to tsunamis for each vulnerability level is determined through a combination of in-depth scenario modelling, historical studies and subject matter expertise. The subsequent recovery from the initial GDP shock is dependent on the socioeconomic resilience assessment of each city. For each resilience level, characteristic recovery profiles are generated through a similar analysis of scenario modelling, historical studies and subject matter expertise.

The consequence analysis for natural disasters was conducted through consultation with subject matter experts. It is supported by a large amount of data and case studies that exist on economic damage from natural disasters, including from EM-DAT, a comprehensive disaster database by the Centre for Research on the Epidemiology of Disasters.

10.7 References The Tsunami Laboratory of the Institute of Computational Mathematics and Mathematical Geophysics URL: http://tsun.sscc.ru/tgi_1.htm

National Oceanic and Atmospheric Administration JetStream Tsunami Locations & Occurrences. URL: http://www.weather.gov/jetstream/locations

11 Flood

11.1 Threat Description Flood risk for a city arises from three main causes: • Riverine Flooding: River bursts its banks from a water load from a catchment area upstream receiving exceptional precipitation or snow-melt • Coastal Flooding and Storm Surge: Sea water is driven onshore by high winds, high tides, and low pressure weather systems • Flash Flood: Localised high precipitation overloads drainage system of city, causing flooding in the urban area Some cities experience several of these flood types. Detailed flood risk assessment has not been carried out for each city, but each city is categorized using flood hazard analysis by United Nations Environmental Program and the Global Archive Map of Extreme Flood Events of Dartmouth Flood Observatory. Cities are categorized by their locations on storm surge-prone coastlines, historical riverine flood events, and past incidences of extensive flash floods. Geographical mapping shows cities on a hydrology mapping of major river systems. Cities are assigned the likelihood of experiencing characteristic flood scenarios and the GDP losses that would result.

11.2 Mapping the Threat Most flood risk mapping is carried out at a detailed level by individual jurisdictions, sometimes by national authorities, but there are few authoritative analyses that have been published that measure the flood hazard for cities worldwide on a consistent basis. We categorize the flood risk of major cities around the world using a number of sources, including: UNEP/DEWA Global estimated risk index for flood hazard, for the Global Assessment Report on Risk Reduction (GAR)40, Dartmouth Flood

40 http://preview.grid.unep.ch/index.php?preview=data&events=floods&evcat=1&lang=eng

Cambridge Centre for Risk Studies 42 Cambridge Global Risk Outlook 2018 Threat Profiles

Observatory Global Archive Map of Extreme Flood Events41, and National Oceanographic Aeronautical Administration Laboratory for Satellite Altimetry and Sea Level Trends42.

These provide first order flood threat assessments, which have been refined with individual city analysis for the cities flagged with significant flood risk, including

• History of major floods and precedents of past flood catastrophes43 • Cities identified as having coastal flood threat have been investigated using Google Earth satellite imagery to identify coastal morphology and topography of the city and the exposure of population centres, the downtown city centre, and economic concentrations to potential coastal floods • Cities identified as being at threat of riverine flooding reviewed using Google Earth satellite imagery to identify major waterways through the city, relative to economic and population exposures • Identifying cities with a threat of storm surge coastal flooding and potential for intense rainfall flash flooding from tropical windstorm events44 • Identifying cities at risk from flash flooding in regions of heavy and intense precipitation, such as monsoon regions45

The assumption is made that most major cities in advanced economies with recurrent riverine or coastal floods have significant flood defences in place.

Cities are given a Threat Assessment Grading as:

Threat Assessment Grade No. of Examples Cities A High Threat of Flood 37 Bangkok, Dhaka, Houston, Nanjing B Moderately High Threat of Flood 57 Bangalore, London, Sao Paolo C Moderate Threat of Flood 82 Berlin, Copenhagen, Mexico City, New York D Moderately Low Threat of Flood 88 Cairo, Denver, Managua E Low Threat of Flood 15 Belgrade, Riyadh

11.3 Local Impact Severity Descriptions The most significant metric of flood severity is water depth, although the velocity of the water flow can be a significant factor in determining the destructive impact of a flood. The level and type of contamination can increase the costs and impacts of a flood catastrophe.

Each city is analysed for the GDP impact and likelihood of experiencing the following characteristic flood scenarios:

LIS Description FL1 10% of city affected by flooding, reaching 1m depth in parts, low velocity water, 3 months recovery period (e.g. Superstorm Sandy New York) FL2 25% of city area affected by flood waters that reach over 3m depth (more than one storey) in parts; Moderate velocity flowing water moderately contaminated. FL3 Over 50% of city land area affected by flooding, reaching more than two storeys in parts, high velocity destructive water flows and highly polluted waters

41 http://www.dartmouth.edu/~floods/Archives/index.html 42 https://www.star.nesdis.noaa.gov/sod/lsa/SeaLevelRise/slr/mssh_2014-1993_300.png 43 EM-DAT. http://www.emdat.be/database 44 EM-DAT. http://www.emdat.be/database 45 EM-DAT. http://www.emdat.be/database

Cambridge Centre for Risk Studies 43 Cambridge Global Risk Outlook 2018 Threat Profiles

11.4 Quantifying the Threat The likelihood assessment of flood risk was determined by a combination of frequency analysis of historical events from EM-DAT and expert judgement.

Annual Likelihood Return Period Threat Assessment Grade FL1 FL2 FL3 FL1 FL2 FL3 A High Threat of Flood 0.0134 0.0045 0.0013 75 224 747 B Moderately High Threat 0.0080 0.0030 0.0009 125 336 1120 of Flood C Moderate Threat of Flood 0.0067 0.0022 0.0007 149 448 1494

D Moderately Low Threat 0.0045 0 0 224 of Flood E Low Threat of Flood 0.0027 0 0 373

11.5 Vulnerability Assessment We characterise the vulnerability of the economy of a city to being flooded by the kind of economic activities that constitutes its principal outputs. Case studies of flood events show that the more advanced, service-driven economies are more disrupted by floods – they rely on power and communications technology that are vulnerable to flood waters, and are impacted by dislocation of the workforce through transportation failures.

Sector Classification Flood Vulnerability Examples A: Service-Dominated Economy (more 5: Economy Very Highly Vulnerable Paris, Dallas, Tokyo than 75% of economy based on services) to Flood Disruption B: Service-Oriented Economy (67-75% of 5: Economy Very Highly Vulnerable Lisbon, Auckland, Vienna economy based on services) to Flood Disruption C: Service with Industry (service >50% 4 Economy Highly Vulnerable to Istanbul, Delhi, Buenos Aires and Industry >25%) Flood Disruption D: Service-Industrial (Service >50%, 4 Economy Highly Vulnerable to Lima, Yekaterinburg, Oslo Industrial >33%) Flood Disruption E: Service with Industrial/Ag Mix 3 Economy Moderately Vulnerable Nairobi, Karachi, Kampala (Service >50%, Industry and Ag 15-30%) to Flood Disruption F: Industry with Service (Industry and 3 Economy Moderately Vulnerable Chengdu, Yerevan, Jakarta Service both over 33%) to Flood Disruption G: Industrial-Oriented Economy 2 Economy Low Vulnerability to Doha, Tripoli (Industrial >50%, Service 20-50%) Flood Disruption H: Agriculture with Industry & Service 1 Economy Very Low Vulnerability Kano, Accra (Ag >30%, Services >30%) to Flood Disruption Source: CCRS Analysis, Oxford Economics GEM Sector Contribution to GDP

11.6 Consequence Analysis The initial impact to GDP due to flooding for each vulnerability level is determined through a combination of in-depth scenario modelling, historical studies and subject matter expertise. The subsequent recovery from the initial GDP shock is dependent on the socioeconomic resilience assessment of each city. For each resilience level, characteristic recovery profiles are generated through a similar analysis of scenario modelling, historical studies and subject matter expertise.

The consequence analysis for natural disasters was conducted through consultation with subject matter experts. It is supported by a large amount of data and case studies that exist on economic damage from natural disasters, including from EM-DAT, a comprehensive disaster database by the Centre for Research on the Epidemiology of Disasters.

Cambridge Centre for Risk Studies 44 Cambridge Global Risk Outlook 2018 Threat Profiles

11.7 References Peduzzi, P. Flood events 1999-2007. UNEP/DEWA/GRID-Europe. Geneva, Switzerland. URL: http://preview.grid.unep.ch/index.php?preview=data&events=floods&evcat=1&lang=eng

Brakenridge, G.R. and Anderson, E. (2004). Global Active Archive of Large Flood Events. Dartmouth Flood Observatory. http://www.dartmouth.edu/~floods/Archives/index.html

D. Guha-Sapir, R. Below, Ph. Hoyois - EM-DAT: The CRED/OFDA International Disaster Database – www.emdat.be – Université Catholique de Louvain – Brussels – Belgium.

12 Volcano

12.1 Threat Description Volcanic eruption threatens the economic activity of cities mainly through ash clouds, which can cause disruption a long way from the volcano itself. The database of the Global Volcanism Program of the Smithsonian Institution provides the location and eruption history of the volcanoes of the world. There are approximately 1500 active volcanoes on land in the world, with the potential to erupt. A volcanic eruption disrupts air travel in the vicinity, deposits ash over a large area with the capability of burying and damaging property, and close to the eruption can destroy with pyroclastic flows, rock bombs, lava and lahar flows. The magnitude of a volcanic eruption is measured on the Volcanic Explosivity Index - a scale from 0 to 8: VEI Ejecta Eruption Description Plume Frequency Examples Occurrences in volume Classifica Height of last 10,000 tion Eruption years* 0 < 10,000 Hawaiian Effusive < 100 m Persistent Kilauea, Many m³ An outpouring of Piton de la lava on the ground Fournaise (as compared with eruptions of ash into the air)

1 > 10,000 Hawaiian/ Gentle 100– Daily Nyiragongo Many m³ Strombolia Low-level, small to 1000 m (2002) n medium volume

2 > Strombolia Explosive 1–5 km Weekly Ruapehu, 3,477 1,000,00 n/Vulcania Dense cloud of ash New Zealand 0 m³ n and gases with (1971), volcanic bombs (2- Mount 3 meters in Sinabung diameter) (2010)

3 > Vulcanian/ Severe 3–15 km Few months Soufriere 868 10,000,0 Pelean Glowing avalanche Hills (1995), 00 m³ of hot ash and Nevado del pyroclastic flows Ruiz, Colombia (1985)

4 > 0.1 Pelean/ Cataclysmic 10–25 ≥ 1 yr Mount Pelee, 421 km³ /Plinian Columns of gas and km West Indies ash (1902), Extends to Eyjafyallajok stratosphere ull (2010)

5 > 1 km³ Plinian Paroxysmal 20–35 ≥ 10 yrs Mount 166 km Vesuvius, Mount St. Helens (1980)

6 > 10 km³ Plinian/Ult Colossal > 30 km ≥ 100 yrs Krakatoa, 51 ra-Plinian Indonesia (1883), Mount Pinatubo,

Cambridge Centre for Risk Studies 45 Cambridge Global Risk Outlook 2018 Threat Profiles

Philippines (1991)

7 > 100 Ultra- Super-colossal > 40 km ≥ 1,000 yrs Tambora 5 (+2 suspected) km³ Plinian (1815)

8 > 1,000 Supervolca Mega-colossal > 50 km ≥ 10,000 yrs Yellowstone 0 km³ nic (Pleistocene) Source: Volcanic Explosivity Index. Center for Educational Technologies: Exploring the Environment. http://ete.cet.edu/gcc/?/volcanoes_explosivity/

12.2 Mapping the Threat Our analysis does not assess the volcanic hazard of individual volcanoes but categorizes them by eruption history and potential for large future eruptions of Volcanic Explosivity Index of 4 to 7. Volcanoes within 1000 km of each city are considered, and their historical rates of eruption are analysed to estimate average eruption rates. This mapping shows a 500km radius around each active volcano, a typical extent of a significant ash fallout from a VEI 5 eruption. Each city is analysed for the GDP impact and likelihood of experiencing local impact severity scenarios. We use the Smithsonian Global Volcanism Program database46 of active volcanoes and plot their locations on a map, and assess the distance to each city to assess the risk of a volcanic eruption causing disruption to a city’s economy. The distance is used to estimate the potential for that city to be impacted by a volcano. Volcanic effects are chiefly the depth of volcanic ash deposited or the potential for pyroclastic flows to cause destruction. The Global Volcanism Program defines an active volcano as one that has erupted since the last ice age (i.e., in the past ~10,000 years). CCRS identifies volcanos that have erupted: • Geologically Active: Within Holocene (11,000 years) • Historically active: Within past 2000 years (AD) • Recently Active: Within past 50 years (since 1964) • Currently Active: Since 2000 For each city, we identify the proximity of active volcanoes, and the number of volcanoes within a 100 and 500 km radius, categorized by volcanoes that have erupted in all periods from 2000 AD. We count the number of known volcanic eruptions of each VEI index level from the known volcanoes in the vicinity of the city using The Large Magnitude Explosive Volcanic Eruptions database (LaMEVE).47 From this analysis, we categorize each city by its proximity to active volcanoes with the capability of large magnitude eruptions, VEI 5 and above, as follows:

Very Historically Recently No. of Threat Assessment Grade Recently Cities Examples Active Active Active A Close to Many Historically Quito, San Active Volcano with Within Within Within Salvador 5 Medium Likelihood of Large 100km 100km 100km Eruption B Close to Historically Within Within Within Manila, 8 Active Volcano with 100km 100km 500km Santiago

46 Global Volcanism Program, 2013. Volcanoes of the World, v. 4.6.3. Venzke, E (ed.). Smithsonian Institution. Downloaded 12 Dec 2017. http://dx.doi.org/10.5479/si.GVP.VOTW4-2013 47 The Large Magnitude Explosive Volcanic Eruptions database : http://www.bgs.ac.uk/vogripa/searchVOGRIPA.cfc?method=searchForm

Cambridge Centre for Risk Studies 46 Cambridge Global Risk Outlook 2018 Threat Profiles

Medium Likelihood of Large Eruption C Close to Historically Naples, Active Volcano with Some Within Within Within Nairobi 7 Likelihood of Large 100km 500km 500km Eruption D In Region of Historically Seattle, Active Volcano with Within Within Within Tokyo, Busan 19 Medium Likelihood of 500km 500km 500km Major Eruption E Distant volcanos with Yangon, Within Within Within historical activity of large 32 Bangkok, 1000km 1000km 1000km eruptions Ningbo F No recent volcanic activity San More than More than in region N/A 208 Francisco, 1000km 1000km Hyderabad

12.3 Local Impact Severities Each city is analysed for the GDP impact and likelihood of experiencing the following characteristic volcano eruption scenarios:

LIS Description VE1 Ashcloud shuts city for extended period, and covers it with several centimeters of ash, preventing air travel, road traffic, port functions, and normal business activity. VE2 Ashcloud covers city to 1m depth, entailing lengthy recovery process VE3 Parts of city impacted by direct effects of volcanic eruption (pyroclastic gases, lahar flows etc.). City evacuated and population not allowed to return for some time.

12.4 Quantifying the Threat We use historical data from the Smithsonian Institute to determine the average likelihood of volcanic eruptions of each VEI level.

VEI Events in Average Average 10,000 Return Frequency Years Period 2 3477 3 0.3477 3 868 12 0.0868 4 278 36 0.0278 5 84 119 0.0084 6 39 256 0.0039 7 4 2500 0.0004 8 Source: Smithsonian Institution Global Volcanism Program via Geoscience News and Information. Volcanic Explosivity Index. http://geology.com/stories/13/volcanic- explosivity-index/ This data is taken into context with historical precedents from three major eruptions we chose as representative scenarios of each magnitude of eruption. These scenarios were used to identify the depth of ash deposit at various distances such that we could determine the likelihood of cities of differing distances from volcanos will be affected.

• VEI 5: Eruption of Mt St Helen, Washington State, USA, 1980. http://pubs.usgs.gov/gip/msh/ash.html • VEI 6: Mount Pinatubo, Philippines, 1990 • VEI 7: Tambora, Indonesia 1815

Cambridge Centre for Risk Studies 47 Cambridge Global Risk Outlook 2018 Threat Profiles

For each of these we obtained and analysed the isopach maps of the depth of ashfall resulting from the eruption. Isopachs are ellipsoidal, determined by the prevailing wind at the time of the eruption. We assess the radius from the centre of the eruption and measure the proportion of the radius band that was covered by the ash of a certain depth. This is taken to represent the likelihood that a future eruption of this magnitude would deposit ash of that depth on a city that was at that distance away from the volcano in any direction. We assume that wind direction is random and that on the day of the eruption, the wind could take the ash plume in any direction.

The analysis provides the following matrix of likelihood to suffer each of these LIS levels with distance from an eruption of VEI magnitude, as follows:

Radius of Land (Km from eruption) Covered with

LIS VE3 LIS VE2 LIS VE1 (Pyroclastics and Deep Ash) (1m of ash) (>1 cm of ash) 90% 50% 10% 90% 50% 10% 90% 50% 10% covered covered covered covered covered covered covered covered covered VEI 5 eruption 5 10 20 15 30 40 25 100 400 e.g. Mt St Helens VEI 6 eruption 10 15 25 30 50 75 100 250 750 e.g. Pinatubo Philippines 1990 VEI 7 eruption 30 40 50 90 120 150 300 500 1000 e.g. Tambora, Indonesia, 1815 Source: CCRS and CAR analysis This results in the following probability levels for experiencing LIS levels in cities categorized with TAG ratings as follows:

Annual Likelihood Return Period Threat Assessment Grade VE1 VE2 VE3 VE1 VE2 VE3 A Close to Many Historically Active Volcano 0.01 0.002 0.0005 100 500 2000 with Medium Likelihood of Large Eruption B Close to Historically Active Volcano with 0.005 0.001 0.00025 200 1000 4000 Medium Likelihood of Large Eruption C Close to Historically Active Volcano with Some 0.003 0.0008 0.0002 300 1200 5000 Likelihood of Large Eruption D In Region of Historically Active Volcano with 0.002 0.0004 0.0001 500 2500 10000 Medium Likelihood of Major Eruption E Distant volcanos with historical activity of large 0.00025 0.00005 0.00002 4000 20000 50000 eruptions F No recent volcanic 0 0 0 activity in region

12.5 Vulnerability Assessment The economic vulnerability of countries and cities to volcanic eruptions is assumed to be related to its physical vulnerability, i.e. the quality of buildings and infrastructure. Each city is therefore assigned to a vulnerability level based off the earthquake vulnerability analysis (see chapter 8). The initial shocks for each volcano vulnerability level is however specific to the volcano consequence analysis.

Cambridge Centre for Risk Studies 48 Cambridge Global Risk Outlook 2018 Threat Profiles

Physical No. of Examples Vulnerability Countries 1 Very Strong 4 Chile, New Zealand, Japan, United States 2 Strong 24 Australia, Canada, Germany, Portugal 3 Moderate 35 Argentina, Bulgaria, Tunisia, Venezuela 4 Weak 28 Bangladesh, Libya, Thailand 5 Very Weak 16 Angola, Sudan, Zimbabwe

12.6 Consequence Analysis The initial impact to GDP due to volcanic activity for each vulnerability level is determined through a combination of in-depth scenario modelling, historical studies and subject matter expertise. The subsequent recovery from the initial GDP shock is dependent on the socioeconomic resilience assessment of each city. For each resilience level, characteristic recovery profiles are generated through a similar analysis of scenario modelling, historical studies and subject matter expertise.

The consequence analysis for natural disasters was conducted through consultation with subject matter experts. It is supported by a large amount of data and case studies that exist on economic damage from natural disasters, including from EM-DAT, a comprehensive disaster database by the Centre for Research on the Epidemiology of Disasters.

12.7 References Global Volcanism Program. (2013). Volcanoes of the World, v. 4.6.3. Venzke, E (ed.). Smithsonian Institution. Downloaded 12 Dec 2017. http://dx.doi.org/10.5479/si.GVP.VOTW4-2013

The Large Magnitude Explosive Volcanic Eruptions database : http://www.bgs.ac.uk/vogripa/searchVOGRIPA.cfc?method=searchForm

Watson, John. (1997) Ash eruption and fallout. USGS. URL: https://pubs.usgs.gov/gip/msh/ash.html

13 Drought

13.1 Threat Description Droughts and water shortages can have significant effects on local economies, including agricultural losses and reduction in output of manufacturing processes. Extended droughts can also cause significant social disruption as populations vie for limited vital resources.

Drought events are classified here as occurring from rainfall deficits in comparison to the “expected” or normal amount of precipitation. Worldwide studies of drought risk by World Bank and historical drought events catalogued by United Nations Environmental Program, as depicted here, provide data on the frequency and severity of droughts. A worldwide drought risk map from The World Bank is presented below. We recognize that as a result of climate change, these maps are likely to change.

Cambridge Centre for Risk Studies 49 Cambridge Global Risk Outlook 2018 Threat Profiles

Source: Global Drought Hazard Distribution 1980-2000. http://www.climatechange- foodsecurity.org/uploads/global_drought_hazard_distribution.png A metric of the severity of droughts is the Palmer Drought Severity Index, or PDSI. Characteristic scenarios use the Palmer Drought Index, as interpreted by US National Integrated Drought Information System, to assess the potential impact on each city’s GDP, and probability of occurrence. Category Description Possible Impacts Palmer Drought Index D0 Abnormally Going into drought: short-term dryness slowing -1.0 to -1.9 Dry planting, growth of crops or pastures. Coming out of drought: some lingering water deficits; pastures or crops not fully recovered D1 Moderate Some damage to crops, pastures; streams, reservoirs, -2.0 to -2.9 Drought or wells low, some water shortages developing or imminent; voluntary water-use restrictions requested D2 Severe Crop or pasture losses likely; water shortages -3.0 to -3.9 Drought common; water restrictions imposed D3 Extreme Major crop/pasture losses; widespread water -4.0 to -4.9 Drought shortages or restrictions D4 Exceptional Exceptional and widespread crop/pasture losses; -5.0 or less Drought shortages of water in reservoirs, streams, and wells creating water emergencies Source: Drought Classification. The National Drought Mitigation Center. University of Nebraska-Lincoln, United States Department of Agriculture, and the National Oceanic and Atmospheric Administration. URL: http://droughtmonitor.unl.edu/AboutUSDM/DroughtClassification.aspx

13.2 Mapping the Threat We primarily used the Global Risk Data Platform of Drought events from 1980-2001 from UNEP/DEWA/GRID-Europe for the threat assessment grading of the cities. This database contains annual global drought events with spatial data, allowing us to determine which cities were affected. The EM-DAT: The OFDA/CRED International Disaster Database of drought events was used as a secondary source for comparison. This data contains country incidence level of droughts from 1900.

Cambridge Centre for Risk Studies 50 Cambridge Global Risk Outlook 2018 Threat Profiles

Regions which are climactically expected to experience long periods without rainfall do not necessarily have a higher risk of drought. The risk assessment is concerned with the economic impact of drought in areas for which precipitation is expected. Threat Assessment Grade No. of Drought No. of Examples Events (1980-2001) Cities A High Threat of Drought >2 62 Denver, Incheon, Tehran

B Moderate Threat of Drought 1-2 102 Beijing, Dhaka, Naples

C Low Threat of Drought 0 115 Indianopolis, Melbourne, Vilnius

13.3 Local Impact Severity Each city is analysed for the GDP impact and likelihood of experiencing the following characteristic drought scenarios:

LIS Description DR1 D2 'Severe Drought': drought causes water consumption restrictions for that city for 6 months, resulting in water rationing for businesses and residential. Water prioritized for industry, agriculture and emergency provision DR2 D3 'Extreme Drought' Three successive seasons of record levels of below average rainfall results in major water shortages for several years DR3 D4 'Exceptional Drought' sustained for multiple years. Major change in precipitation patterns causes extended drought, which results in severe water consumption restrictions for 5 years

13.4 Quantifying the Threat The likelihood of cities in each threat assessment grade to experience of droughts of each local impact severity are below. These probabilities were estimated from the UNEP Global Risk Data Platform and EM-DAT data.

Annual Likelihood Return Period Threat Assessment DR1 DR2 DR3 DR1 DR2 DR3 Grade A High Threat of 0.05 0.02 0.01 20 50 200 Drought B Moderate Threat of 0.02 0.01 0.00 50 100 500 Drought C Low Threat of 0 0 0 Drought

13.5 Vulnerability Assessment The extent to which each city’s economy is affected by drought is determined by the sectoral composition of its GDP. Droughts are assumed to have a milder impact on service-dominated and service-oriented economies as compared to industry and agriculture oriented economies which rely more on water for operations. The drought vulnerability classification is based off the same analysis as flood vulnerability.

Sector Classification Drought Vulnerability Examples A: Service-Dominated Economy (more Paris, Dallas, Tokyo than 75% of economy based on services) 1 Low Drought B: Service-Oriented Economy (67-75% of Vulnerability for Economy Lisbon, Auckland, Vienna economy based on services)

Cambridge Centre for Risk Studies 51 Cambridge Global Risk Outlook 2018 Threat Profiles

C: Service with Industry (service >50% Istanbul, Delhi, Buenos and Industry >25%) Aires D: Service-Industrial (Service >50%, Lima, Yekaterinburg, Oslo Industrial >33%) E: Service with Industrial/Ag Mix Nairobi, Karachi, (Service >50%, Industry and Ag 15-30%) 2 High Drought Kampala F: Industry with Service (Industry and Vulnerability for Economy Chengdu, Yerevan, Service both over 33%) Jakarta G: Industrial-Oriented Economy Doha, Tripoli (Industrial >50%, Service 20-50%) H: Agriculture with Industry & Service Kano, Accra (Ag >30%, Services >30%) Source: CCRS Analysis, Oxford Economics GEM Sector Contribution to GDP

13.6 Consequence Analysis The initial impact to GDP due to droughts for each vulnerability level is determined through a combination of in-depth scenario modelling, historical studies and subject matter expertise. The subsequent recovery from the initial GDP shock is dependent on the socioeconomic resilience assessment of each city. For each resilience level, characteristic recovery profiles are generated through a similar analysis of scenario modelling, historical studies and subject matter expertise.

The consequence analysis for natural disasters was conducted through consultation with subject matter experts. It is supported by a large amount of data and case studies that exist on economic damage from natural disasters, including from EM-DAT, a comprehensive disaster database by the Centre for Research on the Epidemiology of Disasters.

13.7 References Peduzzi, Pascal. Droughts events 1980-2001. UNEP/DEWA/GRID-Europe. Geneva, Switzerland. URL: http://preview.grid.unep.ch/index.php?preview=data&events=droughts&evcat=1&lang=eng

D. Guha-Sapir, R. Below, Ph. Hoyois - EM-DAT: The CRED/OFDA International Disaster Database – www.emdat.be – Université Catholique de Louvain – Brussels – Belgium.

14 Freeze

14.1 Threat Description Extreme temperature events are most disruptive in climatic regions where they occur only rarely and cities are not well prepared, Extreme freeze events in temperate climatic regions disrupt transport, close airports and ports, and damage infrastructure. Extended periods of extreme cold weather have caused severe disruptions and significant economic losses to developed economies in northern latitudes, such as United Kingdom and other European and North American economies. International trading networks rely on clear and timely deliveries via road, rail, sea, and air. Extreme weather limits and disrupts the flow of international trade, creating choke points that impact the global network. The Centre for Risk Studies published a report on Freeze events and the many consequences of extended periods of extremely cold weather, including human, agricultural, infrastructural, commercial and social. 48 The report also provides a catalogue of historical events, detailing the various solar, volcanic, and climatological phenomena causing the freeze, with some detailed examples illustrating the impact caused in each case. Finally, a magnitude scale is provided allowing for estimation of impact based on the number of ‘freezing’ degree days.

48 CCRS. Profile of a Macro-Catastrophe Threat Type: Freeze. URL: http://cambridgeriskframework.com/getdocument/6

Cambridge Centre for Risk Studies 52 Cambridge Global Risk Outlook 2018 Threat Profiles

14.2 Mapping the Threat Freeze threat assessment of cities is based on their Köppen–Geiger climate zones, as presented in the map below.

Source: Kottek, M., J. Grieser, C. Beck, B. Rudolf, and F. Rubel, 2006: World Map of the Köppen-Geiger climate classification updated. Meteorol. Z., 15, 259-263. DOI: 10.1127/0941-2948/2006/0130. The Köppen–Geiger climate zones are further simplified to classify each city by their Threat Assessment Grade:

Threat Assessment Grade No. of Examples Cities A High Threat of Freeze Event 60 Almaty, Chicago, Stockholm B Medium Threat of Freeze Event 59 Brussels, London, Shenzhen C Low or Minimal Threat of Freeze Event 160 Capetown, Santiago, Wuxi

14.3 Local Impact Severity definitions The local impact severity events are defined using a degree day concept: the number of degrees below freezing, combined with the number of days that the freeze persists for.

Each city is analysed for the GDP impact and likelihood of experiencing the following characteristic freeze scenarios:

LIS Description FR1 Freeze of up to 5 deg below 0 deg Celsius for 3 weeks (-20-100 Degree-days) with some snow and ice, moderate winds FR2 Freeze of up to 10 deg below 0 deg Celsius for 8 weeks, combined with deep snow and high winds FR3 Freeze of up to 20 degrees below 0 deg Celsius for 12 weeks, combined with heavy snow and severe ice loads periodically

Cambridge Centre for Risk Studies 53 Cambridge Global Risk Outlook 2018 Threat Profiles

14.4 Quantifying the Threat The likelihood assessment of Freeze and Heatwave (extreme temperature) events were mostly conducted by a consultant of CAR, Victoria Le.49 Temperature records and historical weather accounts, including those compiled by World Health Organization, EM-DAT, and US Centres for Disease Control were analysed to provide the spatial distribution and likelihood assessments.

Source: Lee, W.V. Nat Hazards (2014) 70: 1453. https://doi.org/10.1007/s11069-013- 0884-7

Annual Likelihood Return Period

FR1 FR2 FR3 FR1 FR2 FR3 A High Threat of Freeze Event 0.105 0.020 0.004 10 50 250 B Medium Threat of Freeze Event 0.051 0.005 0.002 20 200 500 C Low or Minimal Threat of 0 0 0 Freeze Event

14.5 Vulnerability Assessment At this stage of model development, cities are currently not differentiated for vulnerability to freeze events. All cities are assumed to receive the same impact to GDP following an event of a given size.

14.6 Consequence Analysis The initial impact to GDP due to freeze events for each vulnerability level is determined through a combination of in-depth scenario modelling, historical studies and subject matter expertise. The subsequent recovery from the initial GDP shock is dependent on the socioeconomic resilience assessment of each city. For each resilience level, characteristic recovery profiles are generated through a similar analysis of scenario modelling, historical studies and subject matter expertise.

The consequence analysis for natural disasters was conducted through consultation with subject matter experts. It is supported by a large amount of data and case studies that exist on economic damage from natural disasters, including from EM-DAT, a comprehensive disaster database by the Centre for Research on the Epidemiology of Disasters.

49 Lee, W.V. Nat Hazards (2014) 70: 1453. https://doi.org/10.1007/s11069-013-0884-7

Cambridge Centre for Risk Studies 54 Cambridge Global Risk Outlook 2018 Threat Profiles

14.7 References Bowman, G; Coburn, A.; Ruffle, S. (2012). Profile of a Macro-Catastrophe Threat Type: Freeze. URL: http://cambridgeriskframework.com/getdocument/6

Lee, W.V. Nat Hazards (2014) 70: 1453. https://doi.org/10.1007/s11069-013-0884-7

Kottek, M., J. Grieser, C. Beck, B. Rudolf, and F. Rubel, 2006: World Map of the Köppen-Geiger climate classification updated. Meteorol. Z., 15, 259-263. DOI: 10.1127/0941-2948/2006/0130.

15 Heatwave

15.1 Threat Description Extreme temperature events are most disruptive in climatic regions where they occur only rarely and cities are not well prepared. Heat waves are less destructive than freeze events but cause social harm, public health issues for the older population and those with health problems, and disruption to many economic processes and activities. Energy demand for air conditioning can outstrip supply and cause systemic failure.

15.2 Mapping the Threat Köppen–Geiger climate zones of the world were mapped and the zones with the potential for high and sustained extremes were categorized, similar to the Freeze threat assessment. Cities are given a TAG assignment of:

Threat Assessment Grade No. of Cities Examples A High Threat of Heatwave Event 32 Athens, Chongqing, New York B Medium Threat of Heatwave Event 120 Birmingham, Hamburg, Toulouse C Low or Minimal Threat of Heatwave 127 Dublin, Quito, Tbilisi Event

15.3 Local Impact Severity The local impact severity events are defined using a degree day concept: the number of degrees above 32° C, combined with the number of days that the heatwave persists for.

Each city is analysed for the GDP impact and likelihood of experiencing the following characteristic heatwave scenarios:

LIS Description HW1 Heatwave of 1-5° above 32° C for 4 weeks (20-100 Degree-days) HW2 Heatwave of 1-8° above 32° C for 8 weeks (50-500 Degree-days) HW3 Heatwave of 1-12° above 32° C for 16 weeks (112-1300 degree-days)

15.4 Quantifying the Threat The likelihood assessment of Freeze and Heatwave (extreme temperature) events were mostly conducted by a consultant of CAR, Victoria Le.50 Temperature records and historical weather accounts, including those compiled by World Health Organization, EM-DAT, and US Centres for Disease Control were analysed to provide the spatial distribution and likelihood assessments.

50 Lee, W.V. Nat Hazards (2014) 70: 1453. https://doi.org/10.1007/s11069-013-0884-7

Cambridge Centre for Risk Studies 55 Cambridge Global Risk Outlook 2018 Threat Profiles

Source: Lee, W.V. Nat Hazards (2014) 70: 1453. https://doi.org/10.1007/s11069-013- 0884-7

Annual Likelihood Return Period HW1 HW2 HW3 HW1 HW2 HW3 A High Threat of Heatwave 0.105 0.020 0.004 10 50 250 Event B Medium Threat of Heatwave 0.051 0.005 0.002 20 200 500 Event C Low or Minimal Threat of 0 0 0 Heatwave Event

15.5 Vulnerability Assessment At this stage of model development, cities are currently not differentiated for vulnerability to heatwave events. All cities are assumed to receive the same impact to GDP following an event of a given size.

15.6 Consequence Analysis The initial impact to GDP due to heatwaves for each vulnerability level is determined through a combination of in-depth scenario modelling, historical studies and subject matter expertise. The subsequent recovery from the initial GDP shock is dependent on the socioeconomic resilience assessment of each city. For each resilience level, characteristic recovery profiles are generated through a similar analysis of scenario modelling, historical studies and subject matter expertise.

The majority of the consequence analysis for natural disasters was conducted through consultation with subject matter experts. It is supported by a large amount of data and case studies that exist on economic damage from natural disasters, including from EM-DAT, a comprehensive disaster database by the Centre for Research on the Epidemiology of Disasters.

15.7 References Lee, W.V. Nat Hazards (2014) 70: 1453. https://doi.org/10.1007/s11069-013-0884-7

Kottek, M., J. Grieser, C. Beck, B. Rudolf, and F. Rubel, 2006: World Map of the Köppen-Geiger climate classification updated. Meteorol. Z., 15, 259-263. DOI: 10.1127/0941-2948/2006/0130.

Cambridge Centre for Risk Studies 56 Cambridge Global Risk Outlook 2018 Threat Profiles

Part D Technology and Space

16 Nuclear Accident

16.1 Threat Description Nuclear power plants have been operating since 1954. They are built as close as possible to the locations of greatest demand, balancing safety with economics. New plants came on line around the world rapidly until the mid-1980s. At their peak in 2002, there were 438 reactors operating. This has declined over time as plants have been decommissioned faster than they have been replaced. This represents some 16,307 operating years of Nuclear reactors. The Centre for Risk Studies has studied the consequences of the nuclear fallouts of Fukushima 2011, Kyshtym, Russia 1957, Windscale UK 1957, and Three Mile Island, USA, 1979. Based on these studies, we assume that a city that affected by a very severe scenario (NP3) of >1Bq/km2 would be effectively abandoned, and that all of its economic output will be lost, although we expect that a lot of that city’s economic output would be displaced – the population would decamp and some or the majority of the economic activity from that city would be resumed elsewhere, re- appropriating capital and recreating the means of production.

The other severities of nuclear pollution also cause disruption to the economy – there is evidence of short term disruption to towns in the Periodic Control Zone around Chernobyl but no population migration or major economic distress. It is however arguable that low levels of radioactive contamination in a city in a Western democracy could have a much more impactful consequence. We have assumed relatively mild economic impact for less severe nuclear fallouts (NP1 and NP2). Thus, the potential for economic losses comes from the potential of heavy localized contamination of a city.

The magnitude of the nuclear power plant accident is measured on the International Nuclear Events Scale (INES). This determines the radius of the footprint of the fallout and how much of the radioactive fuel (‘inventory’) might be released into the atmosphere.

INES Scale Becquerels % of inventory Example released

7 Major accident 100 PBq 10-50% Chernobyl Major release of 1986 radio­active ­material with widespread health and environmental effects requiring implementation of planned and extended ­countermeasures

6 Serious Accident 100 TBq 1-5% Three Mile Significant release of Island 1979 radioactive material likely to require implementation of planned countermeasures

5 Accident with 100 GBq 0.1 - 0.5% Windscale Limited release of wider 1957 radioactive ­material consequences likely to require i­mplementation of some planned­ countermeasures. Several deaths from ­radiation

Cambridge Centre for Risk Studies 57 Cambridge Global Risk Outlook 2018 Threat Profiles

4 Accident with Fuel melt or damage to local consequences fuel ­resulting in more than 0.1% release of core inventory

3 Serious incident Exposure in excess of ten times the statutory annual limit for workers

16.2 Mapping the Threat The World Nuclear Association Information Library provides locations of operational reactors and those that are scheduled to come online within the next decade.51 The mapping of nuclear power plants identifies cities within 250km of operational reactors – the extent of significant fallout from a major core meltdown of grade 7 on the INES Scale. Recorded catalogues of nuclear incidents provide average accidents per year of operation of a reactor, applied to all reactors.

Threat Assessment Grade No. of Examples Cities A Operational NPP within 50 km and multiple NPPs 6 Busan, Taipei within 100km B Operational NPP within 50 km 8 Beijing, Miami C Multiple NPPs within 100km 23 Baltimore, Ningbo, Shenzhen D Operational NPP within 100km 5 Hamburg E Operational NPP within 250km of city 59 Amsterdam, Puebla F No operational NPP within 250km of city 178 Riga, Yangon

16.3 Local Impact Severity Descriptions Fallout from a major core meltdown provides the characteristic scenarios, assessed in terms of radioactive deposit densities, illustrated by distance away from a historical INES 7 event. Each city is analysed for the GDP impact and likelihood of experiencing characteristic nuclear power plant accident scenarios:

LIS Description NP1 City receives radioactive fallout of >0.01Bq/km2 (0.3 Curies of Cs137), similar to within 200km of Chernobyl 1986 or 120km of Fukushima 2011 NP2 City receives radioactive fallout of >0.1Bq/km2 (3 Curies of Cs137) similar to within 70 km of Chernobyl 1986 or 50km of Fukushima 2011 NP3 City receives radioactive fallout of >1Bq/km2 (30 Curies of Cs137) similar to within 30km of Chernobyl INES 7 event in 1986

16.4 Quantifying the Threat The impact of a characteristic scenario – the localized intensity of an event - is graded by the amount of radioactivity that falls at a particular location, measured in Becquerels per square kilometre. Nuclear power plant accidents release different isotypes depending on their fuel mixture and type of accident, some of which are more dangerous than others. The Centre for Risk Studies conducted case

51 World Nuclear Power Reactors & Uranium Requirements. World Nuclear Association. http://www.world- nuclear.org/info/Facts-and-Figures/World-Nuclear-Power-Reactors-and-Uranium-Requirements/

Cambridge Centre for Risk Studies 58 Cambridge Global Risk Outlook 2018 Threat Profiles

studies of several notable nuclear plant accidents including Chernobyl, Windscale, Fukushima and Three Mile Island.

In the Chernobyl accident, the radioactive contamination was chiefly measured in Curies of Cesium 137. The following table provides the impact of the Chernobyl meltdown in 1986, translating Cesium densities into Becquerels:

% of land area affected within radius of

Curies per x10^12 sq km of Bq 50 km 100 km 250 km Cesium 137 Closed 40 curies per sq km of Cesium 40 1.48 50% 5% 2% 137 Permanent Control Zone 15-40 curies per sq km of 15 0.55 30% 10% 5% Cesium 137 Periodic Control Zone 5-15 curies per sq km of Cesium 5 0.185 10% 15% 10% 137 Affected zone 1-15 curies per sq km of Cesium 1 0.037 10% 65% 20% 137

We used this scale to define the magnitudes of events and calibrate the frequency of their historical occurrence.

There have been 6 events of INES 5 and above, an average of 2,718 reactor years per event. The annual probability for an accident of INES 5 and above at a reactor is an average of 0.0004. Some reactors may have higher probabilities than others – possibly due to their age, design, and safety of operation, but for our initial risk estimation we assume that all NPPs have the same probability of failure. Table below shows the probabilities we assume per reactor of a nuclear accident of different magnitudes. INES Scale Number of Cumulative Prob per RP of event events Number of Operation-Year recorded events 7 Major accident 2 2 0.00012 8,154 6 Serious Accident 1 3 0.00018 5,436 5 Accident with wider 4 7 0.00043 2,330 consequences 4 Accident with local 30 37 0.00227 441 consequences 3 Serious incident 62 99 0.00607 165 99

We also assume the following, derived from the analysis of the extent of footprint – the proportion of the land covered by radioactive contamination to different levels of density at various radii from the power plant.

Cambridge Centre for Risk Studies 59 Cambridge Global Risk Outlook 2018 Threat Profiles

Radius of land (kilometres from power plant) covered with

NP3 NP2 NP1

(>1Bq/km2) (>0.1Bq/km2) (>0.01Bq/km2)

90% 50% 10% 90% 50% 10% 90% 50% 10% covered covered covered covered covered covered covered covered covered

INES Grade 5 Event e.g. 0 0 0 1 0 0 5 10 25 Windscale 1957

INES Grade 6 Event e.g. Three 1 2 3 5 15 50 10 30 100 Mile Island 1979

INES Grade 7 Event e.g. 30 50 70 50 70 275 80 150 400 Chernobyl 1986

The resulting likelihoods for each local impact severity and threat assessment are presented here, adjusted from the original analysis using expert judgement:

Return Period Threat Assessment Grade NP1 NP2 NP3 A Operational NPP within 50 km and 8236 5490 3529 multiple NPPs within 100km B Operational NPP within 50 km 9883 12354 49416 C Multiple NPPs within 100km 49416 9883 3801 D Operational NPP within 100km 98833 19766 7602 E Operational NPP within 250km of city 247082 32944 24708

16.5 Vulnerability Assessment At this stage of model development, cities are currently not differentiated for vulnerability to nuclear accident events. All cities are assumed to receive the same impact to GDP following an event of a given size.

16.6 Consequence Analysis The initial impact to GDP due to nuclear power plant accidents for each vulnerability level is determined through a combination of in-depth scenario modelling, historical studies and subject matter expertise. Specific case studies which were analysed include: Chernobyl, Windscale, Fukushima and Three Mile Island. The subsequent recovery from the initial GDP shock is dependent on the socioeconomic resilience assessment of each city. For each resilience level, characteristic recovery profiles are generated through a similar analysis of scenario modelling, historical studies and subject matter expertise.

16.7 References World Nuclear Power Reactors & Uranium Requirements. World Nuclear Association. URL: http://www.world-nuclear.org/info/Facts-and-Figures/World-Nuclear-Power-Reactors-and- Uranium-Requirements/

Map of the Chernobyl Exclusion Zone. Chornobyl Tour. URL: https://chernobyl- tour.com/chernobyl_zone_map_en.html

Cambridge Centre for Risk Studies 60 Cambridge Global Risk Outlook 2018 Threat Profiles

The 'Hot' Nature Created by Sellafield. URL: http://www.lakestay.co.uk/hot.htm

Chernobyl Disaster. https://en.wikipedia.org/wiki/Chernobyl_disaster

Fukushima Nuclear Incidents. URL: http://geospatial.blogs.com/geospatial/fukushima-nuclear- incidents/

17 Power Outage

17.1 Threat Description An extended electrical power outage can cripple economic activity. Power outages result from a wide variety of causes, including other types of threats in this analysis, but the main causes are non- external: accidental damage, power generation shortfalls, operator errors, and component failures. Power disruption statistics for countries from Nation Master Electrical Outage Reports show the fragility of the power grid to shocks and the potential for lengthy and sustained power cuts. Analysis of past power outages in cities indicates the impact of the duration of power loss for the city’s population and economic output. The characteristic scenarios are expressed as numbers of ‘City-Days’ of power loss.

17.2 Mapping the Threat The threat assessment grading for power outages were conducted using counts of days of outages from an aggregate of World Bank Indicator data52 and value lost due to electrical outages as a percent of sales53.

No. of Examples Countries A Very High Threat of Outages (>50 18 Angola, Kenya, Pakistan a year) B High Threat of Outages (10-50 a 22 Chad, India, South Africa year) C Moderate Threat of Outages (1-10 43 Bulgaria, Italy, Turkey a year) D Low Threat of Outages (<1 a year) 24 Australia, Netherlands, United States

17.3 Local Impact Severity Scenarios This threat describes scenarios of wide-scale power outages. To be included, the power outage must conform to all of the following criteria:

• The outage must not be planned by the service provider. • There must be at least 1,000,000 person-hours of disruption. For example, 1,000 people are affected for 1,000 hours (42 days) minimum. If fewer than 1,000 people are affected, the event would not be included regardless of duration. Another example is 1 million people are affected for a minimum of one hour; if the duration is less than one hour, the event would not be included, regardless of number of people.

Each city is analysed for the GDP impact and likelihood of experiencing the following characteristic power outage scenarios:

52 "Countries Compared by Energy > Electrical outages > Days. International Statistics at NationMaster.com", World Development Indicators database. Aggregates compiled by NationMaster. Retrieved from http://www.nationmaster.com/country-info/stats/Energy/Electrical-outages/Days 53 http://data.worldbank.org/indicator/IC.FRM.OUTG.ZS

Cambridge Centre for Risk Studies 61 Cambridge Global Risk Outlook 2018 Threat Profiles

LIS Description PO1 One City-Day of Power Loss (100% of city loses power for 1 day or 50% of city loses power for 2 days etc). PO2 A 5-City-Day event (100% of city loses power for 5 days, 50% of city loses power for 10 days etc) PO3 A 10 City-Day event (100% of city loses power for 10 days)

17.4 Quantifying the Threat We conducted a historical analysis of power outages in New York and their duration, shown below. This analysis was extended to determine the annual likelihood of exceedance for varying city-days of power outage.

New York Power Outages

100% 0.86

90% 0.51 HU Sandy 2012 Derecho Thunderstorm 2012 80% Northeast Blackout 2003 70% 0.73 Lightning Strike Outage 1977 Great Blackout 1965 60%

50%

40%

30% % % of City Without Power

20% Aggregate City-Days of Loss = 10% 2.8

0% 0.32 0 7 14 Number of Days Without Power

100.0% A Very High Threat of Outages (>50 a year) B High Threat of Outages (10-50 a year) C Moderate Threat of Outages (1-10 a year) D Low Threat of Outages (<1 a year)

10.0%

1.0% Annual Exceedance Prob of Annual

0.1% 0.1 1.0 10.0 Aggregate City-Days of lost production

Source: CCRS Analysis

Cambridge Centre for Risk Studies 62 Cambridge Global Risk Outlook 2018 Threat Profiles

Annual Likelihood Return Period Threat Assessment Grade PO1 PO2 PO3 PO1 PO2 PO3 A Very High Threat of Outages (>50 a 0.078 0.015 0.008 13 65 133 year) B High Threat of Outages (10-50 a year) 0.051 0.010 0.005 20 98 200 C Moderate Threat of Outages (1-10 a 0.038 0.008 0.004 26 130 266 year) D Low Threat of Outages (<1 a year) 0.025 0.005 0.003 40 196 400

17.5 Vulnerability Assessment At this stage of model development, cities are currently not differentiated for vulnerability to power outage events. All cities are assumed to receive the same impact to GDP following an event of a given size.

17.6 Consequence Analysis The initial impact to GDP due to power outages for each vulnerability level is determined through a combination of in-depth scenario modelling, historical studies and subject matter expertise. The subsequent recovery from the initial GDP shock is dependent on the socioeconomic resilience assessment of each city. For each resilience level, characteristic recovery profiles are generated through a similar analysis of scenario modelling, historical studies and subject matter expertise.

The primary determinant of the GDP impact is the number of city-days of lost production. Prior work at the Cambridge Centre for Risk Studies on power grid vulnerabilities are the Lloyd’s Business Blackout report at http://cambridgeriskframework.com/getdocument/29 and the Integrated Infrastructure: Cyber Resiliency in Society report at http://cambridgeriskframework.com/getdocument/40.

17.7 References Ruffle, S.; Leverett, E.; Coburn, A.W.; Copic, J.; Kelly, S.; Evan, T. (2015). Stress Test Scenario: Business Blackout: The insurance implications of a cyber attack on the US power grid; Cambridge Risk Framework series; Centre for Risk Studies, University of Cambridge. URL: http://cambridgeriskframework.com/getdocument/29

Kelly, S.; Leverett, E.; Oughton, E. J.; Copic, J.; Thacker, S.; Pant, R.; Pryor, L.; Kassara, G.; Evan, T.; Ruffle, S. J.; Tuveson, M.; Coburn, A. W.; Ralph, D. & Hall, J. W. (2016). Integrated Infrastructure: Cyber Resiliency in Society, Mapping the Consequences of an Interconnected Digital Economy; Cambridge Risk Framework series; Centre for Risk Studies, University of Cambridge. URL: http://cambridgeriskframework.com/getdocument/40

18 Cyber Attack

18.1 Threat Description Economic output from modern service sector economies is heavily reliant on information technology. Cyber attacks, major failures of software, information technology and business applications have increasing potential to significantly impact city GDP. The financial burden for businesses and the economy of malicious cyber attacks is large and growing; Intel Security estimates that cyber crime costs the economy over $400 billion each year. The potential of it do more financial and economic harm can be expected as recent years have seen record breaking incidences of data exfiltration, denial of service, and digital identity theft. 54

54 Risk Management Solutions, Inc. 2017. Cyber Risk Landscape Update, report prepared by RMS in collaboration with the Centre for Risk Studies, University of Cambridge

Cambridge Centre for Risk Studies 63 Cambridge Global Risk Outlook 2018 Threat Profiles

The perpetrators of these attacks are not necessarily motivated by financial gain; increasingly there is a substantial political and espionage element to these attacks. The NotPetya attack in 2017 is believed to be committed by a Russian state-sponsored threat actor that wanted to disrupt public and private industries associated with countries of strategic interest. Russian backed hacking groups are also thought to be behind a large data leak prior to the French election of Emmanuel Macron. The threat posed to different countries by cyber hackers is based off data on cyber incidences and an extensive resource of publications. The vulnerability of each city’s economy to failures of information technology is derived from service sector reliance from OECD national economic data. It is assumed that service sectors are generally more reliant on information technology than other industries such as agriculture and heavy manufacturing. Reference the publications by CCRS on cyber threat as an emerging risk. Publications include:

• 2017 Cyber Risk Landscape http://cambridgeriskframework.com/getdocument/68 • Managing Cyber Accumulation Risk http://cambridgeriskframework.com/getdocument/39 • Integrated Infrastructure: Cyber Resiliency in Society http://cambridgeriskframework.com/getdocument/40 • Lloyd's Business Blackout Scenario http://cambridgeriskframework.com/getdocument/29 • Sybil Logic Bomb Cyber Catastrophe Stress Test Scenario http://cambridgeriskframework.com/getdocument/9

18.2 Mapping the Threat The threat assessment grade assignments are based off extensive background knowledge from internal experts at the Cambridge Centre for Risk Studies. There are few public databases that provide geographical, country or city-level, identification of cyber risk. The threat assessments were determined by historical incidences of events of the three levels of local impact severity.55

Threat Assessment Grade No. of Examples Countries A High Cyber Threat (High Priority Target 23 China, Japan, United for Cyber Attackers) States B Moderate Cyber Threat (Medium Priority 30 Brazil, Malaysia, Saudi Target for Cyber Attackers) Arabia C Low Cyber Threat (Low Priority Target for 54 Algeria, Egypt, Zambia Cyber Attackers)

18.3 Local Impact Severity Descriptions Each city is analysed for the GDP impact and likelihood of experiencing the following characteristic cyber attack scenarios:

LIS Description CY1 Cyber attacks and technology failures increase to such a level that it undermines consumer confidence in technology, weakens productivity of tech sector, and reduces e-commerce CY2 A sustained public cloud outage causes significant downtime to companies' e-commerce platforms and halts access to vital data for continuation of business, resulting in heavy losses to many commercial industries CY3 Cyber attacks on critical infrastructure destroy the power distribution grid and causes power loss in the city for 6 months

55 Please refer to 2017 Cyber Risk Landscape report for background data: http://cambridgeriskframework.com/getdocument/68

Cambridge Centre for Risk Studies 64 Cambridge Global Risk Outlook 2018 Threat Profiles

18.4 Quantifying the Threat The annual likelihood estimates for cyber attacks is difficult to derive as it is an emerging risk with few historical precedents. The landscape is in constant flux: new forms of cyber crime and potential impacts are developing at a fast pace. The return periods for the characteristic threat assessment grades are based off cyber risk expertise at the Centre for Risk Studies. We acknowledge that these estimates will change as more information becomes available and the threat matures.

Annual Likelihood Return Period Threat Assessment Grade CY1 CY2 CY3 CY1 CY2 CY3 A High Cyber Threat (High Priority 0.033 0.005 0.001 30 200 833 Target for Cyber Attackers) B Moderate Cyber Threat (Medium 0.030 0.003 0.001 33 333 1667 Priority Target for Cyber Attackers) C Low Cyber Threat (Low Priority Target 0.010 0.001 0.000 100 1000 5000 for Cyber Attackers)

18.5 Vulnerability Assessment No. of Vulnerability Examples Countries 1 Low Vulnerability to Cyber Attack (Low Afghanistan, Ethiopia 11 dependence on IT for economic productivity) 2 Moderately Low Vulnerability to Cyber Algeria, Libya, United Arab Attack (Moderately Low dependence on IT for 9 Emirates economic productivity) 3 Moderate Vulnerable to Cyber Attack China, Haiti, Thailand (Moderate dependence on IT for economic 21 productivity) 4 High Vulnerability to Cyber Attack (High Bulgaria, Japan, Norway 37 dependence on IT for economic productivity) 5 Very High Vulnerability to Cyber Attack Australia, France, United (Very High dependence on IT for economic 31 States productivity)

18.6 Consequence Analysis The initial impact to GDP due to power outages for each vulnerability level is determined through a combination of in-depth scenario modelling, historical studies and subject matter expertise. The subsequent recovery from the initial GDP shock is dependent on the socioeconomic resilience assessment of each city. For each resilience level, characteristic recovery profiles are generated through a similar analysis of scenario modelling, historical studies and subject matter expertise.

Determining the macroeconomic impacts from these attacks involve estimating the direct impacts, including business interruption and recovery costs, as well as secondary or indirect disruptions to supply chains, other infrastructure and people getting to work. Other factors affected include productivity and consumer confidence.

18.7 References Ruffle, S.; Leverett, E.; Coburn, A.W.; Copic, J.; Kelly, S.; Evan, T. (2015). Stress Test Scenario: Business Blackout: The insurance implications of a cyber attack on the US power grid; Cambridge Risk Framework series and Lloyd’s; Centre for Risk Studies, University of Cambridge. URL: http://cambridgeriskframework.com/getdocument/29

Kelly, S.; Leverett, E.; Oughton, E. J.; Copic, J.; Thacker, S.; Pant, R.; Pryor, L.; Kassara, G.; Evan, T.; Ruffle, S. J.; Tuveson, M.; Coburn, A. W.; Ralph, D. & Hall, J. W. (2016). Integrated Infrastructure: Cyber Resiliency in Society, Mapping the Consequences of an Interconnected Digital Economy;

Cambridge Centre for Risk Studies 65 Cambridge Global Risk Outlook 2018 Threat Profiles

Cambridge Risk Framework series; Centre for Risk Studies, University of Cambridge. URL: http://cambridgeriskframework.com/getdocument/40

CCRS & RMS. (2017). 2017 Cyber Risk Landscape. URL: http://cambridgeriskframework.com/getdocument/68

19 Solar Storm

19.1 Threat Description Solar activity can create geomagnetic and solar radiation storms on earth which can damage electrical circuitry and power transmission systems. Space weather and high energy particles from the sun – a ‘solar storm’ – has the ability to damage our electrical power grids and transmission systems, with the potential to cause lengthy outages that will be very disruptive to our economy. The Centre for Risk Studies has studied the phenomenon of space weather and solar storms disrupting society and the economy, including the Helios Solar Storm publication.56 Space weather can be defined as disturbances of the upper atmosphere and near-Earth space that can disrupt a wide range of technological systems (Hapgood et al. 2012). It can arise from many different types of eruptive phenomena associated with solar activity taking place on the surface of the sun (often referred to as a ‘solar storm’). On average, the Sun’s magnetic activity follows an 11 year solar cycle, with variable minimum and maximum sunspot periods. Solar cycle 24 began in 2008 with minimal sunspot activity until 2010. We are now in the declining phase of the solar cycle where intense activity has previously been more prevalent than other periods (Juusola et al. 2015). The strength and complexity of the Sun’s evolving global magnetic field changes throughout the solar cycle, manifesting as regions of concentrated magnetic field in the photosphere known as sunspots. Through this cycle, the magnetic field in the solar atmosphere alters from a magnetically simple state to a magnetically complex configuration, leading to an increasing number of sunspots (Green and Baker, 2015). While there may be more solar activity during some parts of the solar cycle, solar eruptive phenomena are still the result of a random process. Therefore, there is the potential for this to cause an extreme space weather event affecting Earth at any time. There are three primary forms of solar activity which drive extreme space weather. • Coronal Mass Ejections (CMEs) – CMEs are massive explosions of billions of tonnes of charged particles and magnetic field thrown out into space (Webb and Howard, 2012). • Solar Proton Events (SPEs) – SPEs are a huge increase in energetic particles, mainly of protons but also heavy ions, thrown out into space (Shea and Smart, 2012). They may be related to CMEs and solar flares. • Solar flares – Solar flares are a rapid release of electromagnetic energy previously stored in inductive magnetic fields. Emitted radiation covers most of the electromagnetic spectrum, from radio waves to x- rays (Fletcher et al. 2011). Extreme space weather results from these eruptive solar phenomena.

19.2 Mapping the Threat The threat of an extreme space weather causing disruption to each city is determined by its geomagnetic zoning. Those cities at high geomagnetic latitudes are at highest risk of impact; the magnetic field from solar activity interacts with the Earth’s magnetic field which are strongest near the poles. Modest forms of geomagnetic activity are sometimes seen as auroral lights at high latitudes. In the case of a large coronal mass ejection, these auroral bands can expand toward the equator to lower latitudes, cause rapid changes in the Earth’s magnetic field, producing geomagnetically induced

56 Helios Solar Storm Scenario is available at: http://cambridgeriskframework.com/getdocument/41

Cambridge Centre for Risk Studies 66 Cambridge Global Risk Outlook 2018 Threat Profiles

currents (GICs) which can flow into manmade structures including the electricity transmission network and oil pipelines.

Each city’s threat assessment grade is assigned according to its approximate geomagnetic latitude according to the IGRC model57. Note that geomagnetic latitudes change according to when it is measured and what model is used.

Threat Assessment Grade Geomagnetic No. Examples latitude of Cities A Very High Threat from Solar Storm Events - 55-65 10 Calgary, Reykjvaik, Affected by events of >G2 Tallinn B High Threat from Solar Storm Events - 50-55 25 Boston, Nizhny Affected by events of >G3 Novgorod, Seattle C Moderate Threat from Solar Storm Events - 45-50 27 Berlin, London, Affected by events of >G4 Washington DC D Low Threat from Solar Storm Events - 40-45 27 Atlanta, Milan, Zagreb Affected by events of >G5 E Low Threat from Solar Storm Events 20-40 110 Adana, Brisbane, Taiyuan F Very Low Threat from Solar Storm Events 0-20 80 Dakar, Karachi, Singapore

19.3 Local Impact Severity definitions Each city is analysed for the GDP impact and likelihood of experiencing the following characteristic solar storm scenarios:

LIS Description SS1 NOAA Space Weather Scale for radiation storms level S4 and equivalent to a solar flare of X20. Radiation hazard to passengers and crew in commercial jets at high latitudes (approximately 10 chest x-rays). Satellite systems experience memory device problems and noise on imaging systems, GPS navigation systems prone to error, blackout of HF radio communications. Some low level electrical interference and voltage control problems. 3-5 days of disruption caused. SS2 NOAA Space Weather Scale for radiation storms level S5 and equivalent to a solar flare of X40 (Similar to 'Carrington Event'; Radiation hazard to passengers and crew in commercial jets at high latitudes (approximately 10 chest x-rays). Satellite systems experience memory device problems and noise on imaging systems, GPS navigation systems prone to error, blackout of HF radio communications. Some low level electrical interference and voltage control problems. 3-5 days of disruption caused. SS3 NOAA Space Weather Scale for radiation storms level S6+ (Beyond 5-point NOAA Scale). Estimated effects of a solar flare of X60 - also known as a class Z event. High radiation exposure to passengers and crew in commercial jets at high latitudes (approximately 100 chest x-rays). Satellites rendered useless, GPS navigation systems fail, serious noise on imaging systems. Telecommunication systems fail. Widespread voltage control problems and protective system problems can occur, some grid systems may experience complete collapse or blackouts. Transformers may experience damage. Several weeks of disruption caused before systems back online.

19.4 Quantifying the Threat The observed frequency of solar flares of different magnitudes since 1976 provides extreme value likelihoods for very large solar flares. Their effects on the earth are amplified by the geomagnetic field. Cities located in the strongest geomagnetic latitudes will be worst affected. The study uses the Space Weather Scale for Solar Radiation Storms defined by the US National Oceanic and Atmospheric Administration, to define characteristic scenarios for evaluation of effects on individual cities.

57 VITMO Corrected Geomagnetic Coordinates and IGRF/DGRF Model Parameters. NASA. URL: https://omniweb.gsfc.nasa.gov/vitmo/cgm_vitmo.html

Cambridge Centre for Risk Studies 67 Cambridge Global Risk Outlook 2018 Threat Profiles

Annual Likelihood Return Period Threat Assessment Grade SS1 SS2 SS3 SS1 SS2 SS3 A Very High Threat from Solar Storm 0.067 0.020 0.005 15 51 205 Events - Affected by events of >G2 B High Threat from Solar Storm Events - 0.057 0.017 0.004 18 60 240 Affected by events of >G3 C Moderate Threat from Solar Storm 0.041 0.012 0.003 25 83 333 Events - Affected by events of >G4 D Low Threat from Solar Storm Events - 0.020 0.006 0.002 50 166 666 Affected by events of >G5 E Low Threat from Solar Storm Events 0.013 0.004 0.001 75 250 1000 F Very Low Threat from Solar Storm 0.007 0.002 0.001 150 500 2000 Events

19.5 Vulnerability Assessment The vulnerability assessment of cities affected by solar storms is the same as that of the likelihood assessment. The cities which are more likely to experience a solar storm will also suffer more severe consequences. While there are certain characteristics of cities, e.g. strength of power grid infrastructure, which determine the potential extent of damage, these components have not yet been analysed in detail at this stage of model development.

Threat Assessment Grade Geomagnetic No. Examples latitude of Cities A Very High Threat from Solar Storm Events - 55-65 10 Calgary, Reykjvaik, Affected by events of >G2 Tallinn B High Threat from Solar Storm Events - 50-55 25 Boston, Nizhny Affected by events of >G3 Novgorod, Seattle C Moderate Threat from Solar Storm Events - 45-50 27 Berlin, London, Affected by events of >G4 Washington DC D Low Threat from Solar Storm Events - 40-45 27 Atlanta, Milan, Zagreb Affected by events of >G5 E Low Threat from Solar Storm Events 20-40 110 Adana, Brisbane, Taiyuan F Very Low Threat from Solar Storm Events 0-20 80 Dakar, Karachi, Singapore

19.6 Consequence Analysis The initial impact to GDP due to solar storms for each vulnerability level is determined through a combination of in-depth scenario modelling, historical studies and subject matter expertise. The subsequent recovery from the initial GDP shock is dependent on the socioeconomic resilience assessment of each city. For each resilience level, characteristic recovery profiles are generated through a similar analysis of scenario modelling, historical studies and subject matter expertise.

Estimating the macroeconomic impacts from solar storm events is similar to that for Power Outage, as both analyses involve assessing impacts to the power grid. Additional impacts considered here include loss of radio communications, GPS, and interference to radar systems and satellite operations.

The analysis involves estimating the resulting direct impacts, such as business interruption and recovery costs, as well as secondary or indirect disruptions to supply chains, other infrastructure and people getting to work. Other factors affected include productivity and consumer confidence.

Cambridge Centre for Risk Studies 68 Cambridge Global Risk Outlook 2018 Threat Profiles

19.7 References Oughton, E.; Copic, J.; Skelton, A.; Kesaite, V.; Yeo, Z. Y.; Ruffle, S. J.; Tuveson, M.; Coburn, A. W.; Ralph, D. (2016). Helios Solar Storm Scenario; Cambridge Risk Framework series; Centre for Risk Studies, University of Cambridge. VITMO Corrected Geomagnetic Coordinates and IGRF/DGRF Model Parameters. NASA. URL: https://omniweb.gsfc.nasa.gov/vitmo/cgm_vitmo.html

Cambridge Centre for Risk Studies 69 Cambridge Global Risk Outlook 2018 Threat Profiles

Part E Health and Humanity

20 Human Pandemic

20.1 Threat Description Disease epidemics have been the causes of some of the worst socio-economic shocks throughout human history. At its most extreme, large parts of 14th Century Europe lost a third of its population to the terrible plague of Black Death. Table 21.1 provides a history of some of the most severe epidemics that have affected the world.

Date Name Cause 2012 Middle East Respiratory Syndrome Coronavirus MERS-CoV 2002 Severe Acute Respiratory Syndrome SARS 1981-today Acquired Immunodeficiency Syndrome HIV/AIDS 1918-1922 Russian Typhus Epidemic Typhus 1855-1959 Third Pandemic Bubonic plague 1962-1966 El Tor Cholera Pandemic Cholera 1899-1923 Sixth Cholera Pandemic Cholera 1881-1896 Fifth Cholera Pandemic Cholera 1863-1875 Fourth Cholera Pandemic Cholera 1846-1863 Third Cholera Pandemic Cholera 1826-1837 Second Cholera Pandemic Cholera 1816-1824 Asiatic Cholera Pandemic Cholera 1793; 1690-1878 Yellow fever, U.S. Yellow fever 1775-1782 North American smallpox Smallpox 1679 Great plague of Vienna 1665-1666 Great plague of London 1629-1631 Italian plague/Great Plague of Milan Bubonic plague 16th C Spread of smallpox thru colonization Smallpox 1500-1800 Epidemics throughout Europe Multiple 1577-1579 Following Black Assize 1489 Spanish Siege of Moorish Granada Typhus 1347-1350 Black Death Bubonic plague 639 Plague of Emmaus/Amwas Bubonic plague? 541-750 Plague of Justinian Bubonic plague 251-266 Plague of Cyprian Smallpox or measles? 165-180 Antonine Plague Smallpox or measles? 430 BC Plague of Athens Typhoid/Plague/ Measles?

Table 21.1: Historical Infectious Disease Pandemics58

Disease impacts are not just an ancient historical anomaly – this current generation has had to deal with the impact of HIV/AIDS – a previously unknown disease that medical science could not combat and that has killed 30 million people, many of them wealthy, educated people with access to the best healthcare available in the world. Nature and mankind are engaged in a constant arms race – nature evolves new strains of pathogens to overcome natural defense mechanisms and infect human hosts, and mankind develops antibodies in response, and in modern times has augmented this with medical treatments Twenty known diseases have recently re-emerged or spread geographically. These new outbreaks are of more virulent and drug-resistant forms. At least 30 unknown disease agents for which no cures are available have been identified in human populations in the last few decades, including HIV, Ebola,

58 Sources: Hays, JN, 2005; Little LK, 2006

Cambridge Centre for Risk Studies 70 Cambridge Global Risk Outlook 2018 Threat Profiles

and hepatitis C and E. Infectious disease outbreaks pose a major threat both nationally and internationally. They easily cross borders and can threaten economic and regional stability. Rapid Adaptation of Pathogens

Viruses are exceptionally adaptable organisms. They are constantly undergoing genetic change and can undergo many generations of reproduction in a short period, evolving rapidly. Their adaptation through high mutation rates is partly because their reproductive processes have fewer genetic ‘proof- reading’ checks, particularly RNA viruses. High mutation rates enable random changes to explore vulnerabilities in their human or animal hosts. The adaptation of pathogens to kill them has led to an increasingly worrying phenomenon of antimicrobial resistance (AMR). This is a particular concern in countries that have access to antibiotics but don’t use them properly or administer them unnecessarily. For the sick that are already vulnerable such as the elderly, a combination of antimicrobial resistance to what is a common hospital infection can prove deadly. Growing Reservoirs of Hosts

In addition, the populations of animals that they inhabit and replicate through are increasing rapidly. The global human population has doubled since 1970. Poultry stocks and farmed animal populations have seen massive increases as the developing world demand for protein in their diet has grown. In China alone, the poultry population is estimated to have increased from fewer than one billion birds in 1980 to over 20 billion today. Pigs have increased from 50 million to over 700 million. These and other mammal populations are the reservoirs in which virus mutations take place, finally jumping from one species to another to infect humans. Many of the emergent infection diseases over the past few generations have their origins in the rapidly expanding but poorly regulated agricultural industries of the developing economies, where their close integration with human activity makes it easier for disease outbreaks to transfer from the animal hosts to people. Man-Made Pandemic Risk

Many countries now have sophisticated biological research laboratories handling dangerous pathogens, as biotechnology develops rapidly as a global industry. These laboratories – Biosafety Level 3 and above – are run with high safety standards, but they are complex systems and accidents do happen. There are now at least 42 known laboratories currently working with potential pandemic pathogens (PPPs) – i.e. H5N1 viruses, live versions of the 1918 influenza virus, or the SARS virus. Statistics on the accident record of laboratories show that incidents are rare but significant. Over 5,000 researchers have suffered from some type of laboratory-acquired infections (LAIs) since 1930, and nearly 200 have died. Only a few recorded cases of laboratory accidents have resulted in any kind of epidemic, but one example is the 2007 outbreak of foot-and-mouth disease in cattle in England as a result of a virus escape from the Pirbright BSL-4 research laboratory. The 1977 Russian flu epidemic may have emerged from a laboratory virus escape. In 2012 virologists created a transmissible version of deadly H5N1 in the laboratory – a discovery that caused a scientific controversy over the potential for researchers to accidentally trigger the pandemic they were trying to prevent. After a brief moratorium this ‘HPAI H5N1 gain-of-function’ research recommenced. There is a small but non-zero chance that the next pandemic could be triggered by a laboratory accident. Influenza One of the most rapidly mutating viruses is influenza, a highly contagious RNA virus which has been responsible for some of the most widespread pandemics in recent history, table 2. In an early epidemic in 15th Century Italy, the illness was attributed to “influence of the stars”, hence “influenza”. Influenza has proven very difficult to combat, because it changes so often. Vaccines need to be developed to match the particular strain in circulation – a process that takes several months each time. A new vaccine has to be developed each year for the seasonal flu strain that occurs during winter. Influenza is constantly present in the human population and mutates to a new strain each year, causing a seasonal peak of infection each winter. It is a leading cause of infectious disease-related deaths in most countries around the world. In non- pandemic years influenza typically kills hundreds of thousands of people worldwide. The highest rates of mortality are in the elderly followed by children and those with pre-existing medical problems.

Cambridge Centre for Risk Studies 71 Cambridge Global Risk Outlook 2018 Threat Profiles

Every so often, the gradual mutations of the influenza virus (antigenic drift) give rise to a major genetic change (an antigenic shift or reassortment) that finds a new mechanism to infect humans and evade their immune systems, spreading rapidly through the population to cause widespread illness in a pandemic.

Date Notes Influenza Strain 2009 Mexican Swine Flu H1N1 1977-1978 Russian Flu ‘benign’ pandemic, possibly caused by a lab release H1N1 1968 Hong Kong Flu H3N2 1957-1958 Asian Flu Pandemic H2N2 1918-1919 Spanish Flu ‘The Great Influenza’ H1N1 1889-1893 Russian Flu H3N8 or H2N2 1830-1848 Four influenza epidemics occurring almost continuously 1830 to 1848, possibly originating in China 1788-1790 Initiated a pandemic era, of heightened global influenza activity for almost 20 years 1780-1782 Began in Southeast Asia and spread to Russia and eastward into Europe 1761-1762 Begun in Americas and spread to Europe and around the globe. First pandemic to be scientifically studied. 1729-1730, 1732- First detected in Russia 1733 1580 Swept over the entire globe, spreading east to west from Asia 1557-1558 Asia origin. Highly fatal, and associated with severe complications 1510 First recognizable pandemic. Invaded Europe from Africa.

Table 21.2: Historical Influenza Pandemics59

The threat assessment for human pandemics and emerging infectious disease to business and the economy is largely based off some of the in-depth studies Cambridge Centre for Risk Studies has made such as São Paulo Virus Pandemic Stress Test Scenario at http://cambridgeriskframework.com/getdocument/8, and the Ebola Contingency Scenario at http://cambridgeriskframework.com/getdocument/23.

20.2 Mapping the Threat Pandemics are, by definition, international and they can be expected to spread through all human communities throughout the world when they occur, so all locations and cities will experience a major pandemic event. However, the virulence of a pandemic is slightly higher at its initial outbreak and in the first few generations of the replication of the new virus. Locations where new epidemic outbreaks are most likely are at higher risk of more severe impacts from future pandemics. We identify the locations of higher risk for new outbreaks from a study of the global distribution of the relative risk of an emerging infectious disease event by the Institute of Zoology UK.60 This study identifies locations of past zoonotic outbreaks and similar locations where concentrations of human populations have suffered from novel pathogen infections that jump species from domesticated animals or wildlife.

This mapping is used to grade countries into Threat Assessment Grades:

Threat Assessment Grade No. of Examples Countries A High Threat of Emerging Infectious Diseases 17 Bangladesh, Kenya B Moderately High Threat of Emerging 46 Brazil, Israel, Ukraine Infectious Diseases

59 Taubenberger, JK; Morens, DN. (2009). Pandemic influenza--including a risk assessment of H5N1. Rev Sci Tech. 60 Jones et al., (2007) Global Trends in Emerging Infectious Diseases. Nature. Study by Institute of Zoology, UK, Consortium for Conservation Medicine, New York.

Cambridge Centre for Risk Studies 72 Cambridge Global Risk Outlook 2018 Threat Profiles

C Possible Threat of Emerging Infectious 34 Afghanistan, France, Diseases Singapore D Low Threat of Emerging Infectious Disease 10 Finland, Iceland

20.3 Local Impact Severities Each city is analysed for the GDP impact and likelihood of experiencing the following characteristic infectious disease scenarios:

LIS Description HE1 Localized epidemic of new emergent disease that has case fatality rate of 10% causes public health emergency and fear in population of catching disease, leads to loss of tourism trade (e.g. SARS) HE2 Highly infectious moderate virulence pandemic where influenza virus infects 43% of the population, with a case fatality rate of 0.3% (e.g. Sao Paulo Scenario); anti-microbial resistance seen in hospitals in some countries HE3 Highly virulent pandemic causes high (3%) fatality rate in infected cases

HE1 assumes that if an emerging infectious disease breaks out then it will affect 3 countries and that the cities in that country have a 75% chance of being affected. Countries with a higher threat from emerging infectious disease, with a higher TAG, are more likely to experience these localised epidemics.

HE2 and HE3 assumes that every city in the world experiences infection from the pandemic. The main difference in the impact comes from the vulnerability of the city, which is determined by the healthcare quality being offered in that country. In these two scenarios, we also see anti-microbial resistance in some countries. AMR marginally increases the fatality rate as highly vulnerable patients are susceptible to microbes in the hospital setting, potentially triggering secondary infections that become difficult to treat.

20.4 Quantifying the Threat The annual likelihood estimates for each threat assessment grade and characteristic scenario are derived from a combination of historical frequency analysis and the probabilistic event set from the RMS Infectious Disease Model (IDM) with permission from RMS. This is the same event set that was used to identify the scenario used for an in-depth scenario modelling of the Sao Paolo Virus Pandemic Scenario.61

Annual Likelihood Return Period Threat Assessment Grade HE1 HE2 HE3 HE1 HE2 HE3 A High Threat of Emerging 0.005 0.0075 0.0023 200 133 438 Infectious Diseases B Moderately High Threat of 0.002 0.0075 0.0023 500 133 438 Emerging Infectious Diseases C Possible Threat of Emerging 0.00125 0.0075 0.0023 800 133 438 Infectious Diseases D Low Threat of Emerging 0.001 0.0075 0.0023 1000 133 438 Infectious Disease

20.5 Vulnerability Assessment The rating of each country’s healthcare system is based on the Access to Healthcare Index of the INFORM risk index. This index is an aggregate of factors including physician density, measles

61 Ruffle, S.J.; Bowman, G.; Caccioli, F.; Coburn, A.W.; Kelly, S.; Leslie, B.; Ralph, D. (2014). Stress Test Scenario: São Paulo Virus Pandemic; Cambridge Risk Framework series; Centre for Risk Studies, University of Cambridge. URL: http://cambridgeriskframework.com/getdocument/8

Cambridge Centre for Risk Studies 73 Cambridge Global Risk Outlook 2018 Threat Profiles

immunization coverage, per capita public and private expenditure on healthcare, and maternal mortality ratio. This rating was compared to the World Health Organization report from 2000 which provided one of the first comprehensive global rankings of health systems (but was subsequently disaggregated into multiple health indicators).62 The INFORM index was used as the primary data source as it is a more recent composite index, reflecting significant changes in healthcare quality in some countries in the past two decades.

An additional threat assessment grade 0 was included in the 2018 update to account for countries with very strong healthcare systems but are experiencing heightened levels of antimicrobial resistance (AMR). Adjustments were not made to those countries which exhibit effects of AMR but have lower ranked healthcare systems; it is assumed that the effects of inadequate administration of medication is already embedded in the impact associated with that threat assessment grade.

Relative levels of 4 types of antimicrobial resistance were determined from data in ResistanceMap from The Center for Disease Dynamics, Economics & Policy.63 Countries with Level 1 - Very Strong Healthcare Systems that had high levels of either of the 4 types of AMR were re-classified as Level 0 – Very Strong Healthcare with signs of AMR.

Threat Assessment Grade Access to No. of Examples Healthcare Countries Range 0 Very Strong Healthcare with signs 0-1.3 5 France, Greece, Italy, Spain, of AMR United States 1 Very Strong Healthcare System 0-1.3 20 Australia, Denmark, Sweden 2 Strong Healthcare System 1.4-3.2 25 Argentina, South Korea, Slovenia 3 Moderate Healthcare System 3.3-5.4 28 China, Malaysia, Venezuela 4 Weak Healthcare System 5.5-7.2 18 Bangladesh, Iraq, Zimbabwe 5 Very Weak Healthcare System 7.3-9.9 11 Afghanistan, Haiti, Pakistan

20.6 Consequence Analysis The initial impact to GDP due to human pandemic for each vulnerability level is determined through a combination of in-depth scenario modelling, historical studies and subject matter expertise. The subsequent recovery from the initial GDP shock is dependent on the socioeconomic resilience assessment of each city. For each resilience level, characteristic recovery profiles are generated through a similar analysis of scenario modelling, historical studies and subject matter expertise.

The two key parameters in determining the impact of a pandemic is its transmissibility and virulence. Transmissibility is the how fast the disease spreads through the population and proportion of the population it ultimately infects. Virulence is how severely it makes people sick. There is a trade-off between the two parameters: as virulence increases, more hosts die, people are more scared and reduce their contact rate, thus reducing the infection rate.

Macroeconomic factors affected by a pandemic include labour supply and productivity due to workforce absenteeism and knock-on impacts to supply chains; reduction in demand as discretionary consumption is cut; direct impacts to sectors such as travel, tourism and hospitality; and increases to government expenditure due to increase in emergency response and health care provision.

20.7 References Kelly, S.; Coburn, A.W.; Ebola Contingency Scenario: Analysis of Economic Impact of Upper Bound Ebola Projections for US and Europe - Cambridge ‘Contingency’ Scenario developed for business

62 http://www.who.int/healthinfo/paper30.pdf 63 https://resistancemap.cddep.org/AntibioticResistance.php

Cambridge Centre for Risk Studies 74 Cambridge Global Risk Outlook 2018 Threat Profiles

preparedness planning; Working Paper 2014: 2; Cambridge Risk Framework series; Centre for Risk Studies, University of Cambridge. URL: http://cambridgeriskframework.com/getdocument/23

Ruffle, S.J.; Bowman, G.; Caccioli, F.; Coburn, A.W.; Kelly, S.; Leslie, B.; Ralph, D. (2014). Stress Test Scenario: São Paulo Virus Pandemic; Cambridge Risk Framework series; Centre for Risk Studies, University of Cambridge. URL: http://cambridgeriskframework.com/getdocument/8

Tandon, A.; Murray, C.J.L.; Lauer, J.A.; Evans, D.B. (2000). Measuring Overall Health System Performance for 191 Countries; World Health Organization; GPE Discussion Paper Series: No. 30. URL: http://www.who.int/healthinfo/paper30.pdf

Taubenberger JK; Morens DM. Pandemic influenza--including a risk assessment of H5N1. Rev Sci Tech. 2009 Apr;28(1):187-202. URL: https://www.ncbi.nlm.nih.gov/pubmed/19618626

Jones et al., (2007) Global Trends in Emerging Infectious Diseases. Nature. Study by Institute of Zoology, UK, Consortium for Conservation Medicine, New York.

Hays, JN. (2005). Epidemics and Pandemics: Their Impacts of Human History. Santa Barbara, CA.

Little, L.K. (2006). Plague and the End of Antiquity. Cambridge. Cambridge University Press.

21 Plant Epidemic

21.1 Threat Description One of the major causes of food shortages and price increases is crop disease and harvest failures. The impact of staple crop failures is greatest in those cities where agriculture makes up a high proportion of the economy. Widespread epidemics and harvest failures cause hikes in global food prices will affect all cities. The evolution of this threat is more long-term: plant diseases develop over several years and often can be managed, although this can prove very expensive. Wheat rust is one of the most significant sources of concern for global food security, as wheat makes up 15% of global caloric consumption. This fungal disease can reduce wheat yield by up to 70% in East Africa through to the Middle East.64 Another worrying trend is the reduction of biodiversity as industrial farming practices grow globally, reducing biodiversity and the resilience of plant strains and societies from pests and disease. However, efforts at improving surveillance and development of new fungicides provide optimism against potential disasters.

21.2 Mapping the Threat We used distribution maps of major plant diseases from Plantwise CAB International that show the likelihood of plant diseases occurring in major staple crops, like wheat and rice.65 These images were visually inspected to assign threat assessment grades to each city and country.

64 Chamy, C. “Wheat rust: The fungal disease that threatens to destroy the world crop” The Independent. http://www.independent.co.uk/news/uk/home-news/wheat-rust-the-fungal-disease-that-threatens-to-destroy-the-world- crop-9271485.html 65 http://www.plantwise.org/KnowledgeBank/PWMap.aspx

Cambridge Centre for Risk Studies 75 Cambridge Global Risk Outlook 2018 Threat Profiles

Figure 21.1. Example of Plantwise CAB International database of countries affected by Puccinia recondite (wheat rust)

Threat Assessment Grade No, of Examples Countries A High Threat of Plant Epidemic in 37 Haiti, Malaysia, Taiwan Local Crops B Moderate Risk of Plant Epidemic in 29 Chile, Morocco, United States Local Crops C Low Risk of Plant Epidemic in Local 41 Argentina, Ireland, Saudi Crops Arabia

21.3 Local Impact Severity Descriptions Each city is analysed for the GDP impact and likelihood of experiencing the following characteristic plant epidemic scenarios:

LIS Description PE1 Localized Plant Epidemic affects prices of staple foods in city markets PE2 National plant epidemic affects price of staple foods in city markets PE3 International Plant Epidemic affects price of stable foods in city markets

21.4 Quantifying the Threat The annual likelihood of plant epidemic is largely based off historical frequency analysis of significant plant disease epidemics in addition to expert consultation with the Department of Plant Sciences at the University of Cambridge.

Annual Likelihood Return Period Threat Assessment Grade PE1 PE2 PE2 PE1 PE2 PE2 A High Threat of Plant Epidemic 0.10 0.02 0.01 10 50 100 in Local Crops B Moderate Risk of Plant 0.04 0.01 0.01 25 100 100 Epidemic in Local Crops C Low Risk of Plant Epidemic in 0.01 0.00 0.01 100 500 100 Local Crops

Cambridge Centre for Risk Studies 76 Cambridge Global Risk Outlook 2018 Threat Profiles

21.5 Vulnerability Assessment The vulnerability of city and country GDP to plant epidemics is dependent upon the sectoral composition of the economy. This assessment is based off the same underlying data from the flood and drought vulnerability assessments. Countries that have highly agrarian economies are considered more vulnerable to widespread plant disease while countries that are industry and service-based are relatively less vulnerable.

Sector Classification Vulnerability Examples A: Service-Dominated Economy (more than 75% of economy based on services) B: Service-Oriented Economy (67- 75% of economy based on services) 3 Economy Has Low Argentina, Republic of C: Service with Industry (service Vulnerability to Plant Congo, India, Norway >50% and Industry >25%) Epidemic D: Service-Industrial (Service >50%, Industrial >33%) G: Industrial-Oriented Economy (Industrial >50%, Service 20-50%) E: Service with Industrial/Ag Mix (Service >50%, Industry and Ag 15- 2 Economy Moderately 30%) Vulnerable to Plant China, Egypt, Zambia F: Industry with Service (Industry Epidemic and Service both over 33%) H: Agriculture with Industry & 1 Economy Very Service (Ag >30%, Services >30%) Vulnerable to Plant Afghanistan, Chad, Nigeria Epidemic

21.6 Consequence Analysis The initial impact to GDP due to plant epidemic for each vulnerability level is determined through a combination of in-depth scenario modelling, historical studies and subject matter expertise. The subsequent recovery from the initial GDP shock is dependent on the socioeconomic resilience assessment of each city. For each resilience level, characteristic recovery profiles are generated through a similar analysis of scenario modelling, historical studies and subject matter expertise.

21.7 References CABI Plantwise. URL: https://www.plantwise.org/KnowledgeBank/PWMap.aspx Plant pests and diseases. (2017). FAO. http://www.fao.org/emergencies/emergency-types/plant- pests-and-diseases/en/

Cambridge Centre for Risk Studies 77