Insurance Risk Study - Aon Benfield

Insurance Risk Study Seventh Edition 2012

Empower Results™

Insurance Risk Study

Contents 3 Foreword 4 Global Risk Parameters 6 U.S. Risk Parameters 7 U.S. Reserve Adequacy and Risk 8 Modeling the Underwriting Cycle 9 New Paradigm Optimization 10 Motor Insurance in China 11 A Different View of Risk 18 Global Wind and Earthquake Risk 20 Robust Risk Management 21 Global Crop Insurance Penetration 22 Correlation Considerations 23 Global Correlation between Lines 24 Macroeconomic Correlation 25 Global Market Review 31 Afterword: The Disappearing Risk

About Aon Benfield Aon Benfield, a division of Aon plc (NYSE: AON), is the world’s leading reinsurance intermediary and full-service capital advisor. We empower our clients to better understand, manage and transfer risk through innovative solutions and personalized access to all forms of global reinsurance capital across treaty, facultative and capital markets. As a trusted advocate, we deliver local reach to the world’s markets, an unparalleled investment in innovative analytics, including catastrophe management, actuarial and rating agency advisory. Through our professionals’ expertise and experience, we advise clients in making optimal capital choices that will empower results and improve operational effectiveness for their business. With more than 80 offices in 50 countries, our worldwide client base has access to the broadest portfolio of integrated capital solutions and services. To learn how Aon Benfield helps empower results, please visit aonbenfield.com.

Aon Benfield

Foreword Twenty years ago, on August 24, 1992, Hurricane Andrew made landfall in Homestead, Florida, and fundamentally changed the insurance industry forever. It precipitated the analytic era of insurance risk management, with its focus on computer simulation models and statistical quantification of risk. Results of the last twenty years show the great strengths of the new era, but also some weaknesses. The industry has become both more robust and more fragile. 2005 demonstrated the new robustness: despite four landfalling hurricanes, including Katrina, only seven U.S. companies became impaired in 2005, the second lowest total in the last twenty years. 2001 showed fragility with the surprise World Trade Center terrorist attack, which contributed to the 41 impairments that year, the second highest total. 2011 again challenged the global insurance industry with unexpected, un-modeled losses from the Tōhoku tsunami and Thailand floods. Time and time again we see perils with robust, widely accepted models being more effectively managed through insurance, reinsurance and other mechanisms. There is now broad agreement on the need to model and understand as many insured risks as possible. Since its first edition in April 2006, the Insurance Risk Study has provided a consistent set of global benchmark risk parameters based on empirical data. The Study is designed to put the analysis of liability and non-catastrophic property lines on an equal footing with risks analyzed using catastrophe models and to help expand the universe of effectively modeled risk as far as possible. In this year’s Study we maintain our philosophy of complementing catastrophe models. However, motivated by the fragility caused by un-modeled events, we also report on catastrophic risk — but from a purely empirical point of view. A series of Aon Benfield cartogram maps, starting on page 11, illustrate global exposure to wind, earthquake and flood perils. These cartogram maps are based on up to 100 years of historical data. They form the basis for a robust, consistent, and global complement and backup to traditional catastrophe models. Last year’s tsunami, Thailand floods and U.S. severe weather outbreaks, which were not modeled or only poorly modeled, illustrate the need for such backup risk management. Another addition to the Study this year provides expanded coverage of opportunities in the insurance world. Which parts are growing and which are shrinking? Where are the opportunities to write more profitable business? Aon Benfield can provide its clients with very granular market intelligence to create business plans that are realistic, actionable, fact-based and, above all, profitable. Our approach, from sizing the opportunity through identifying distribution channel dynamics, assessing competitor behavior, and understanding what it takes to compete and win, is illustrated and explained on pages 25 to 30. The Aon Benfield Insurance Risk Study continues to be the industry’s leading set of risk parameters for modeling and benchmarking underwriting risk. It is part of a suite of capabilities that help position Aon Benfield as a true business partner on questions of risk management, growth and operational performance. The Study is a cornerstone of Aon Benfield Analytics’ integrated and comprehensive risk modeling and risk assessment capabilities. Our reinsurance optimization framework, explicitly linking reinsurance and capital and quantifying the resulting volatility reduction, combines client data with risk parameters from the Study to generate a credible assessment of baseline underwriting volatility. Our global risk and capital strategy practice, providing ERM and economic capital modeling services and the ReMetrica® software platform, uses the Study to benchmark risk, quantify capital adequacy and allocate capital to risk drivers. The massive database underlying the Study is supported by more than 400 local professionals within Aon Benfield’s global analytics team who are available to work with you to customize the basic parameters reported in the Study to answer your specific, pressing business questions. All of our work within the analytics team is motivated by client questions. We are grateful to clients who have invited us to share in the task of helping them analyze their most complex business problems. Dynamic and interactive working groups always lead to innovative, and often unexpected, solutions. If you have questions or suggestions for items we could explore in future editions please contact your local Aon Benfield broker or any of the contacts listed on the back page.

3


Global Risk Parameters The 2012 Insurance Risk Study quantifies the systemic risk by line for 48 countries worldwide, representing more than 90 percent of global premium. Systemic risk in the Study is the coefficient of variation of loss ratio for a large book of business. Coefficient of variation (CV) is a commonly used normalized measure of risk defined as the standard deviation divided by the mean. Systemic risk typically comes from non-diversifiable risk sources such as changing market rate adequacy, unknown prospective frequency and severity trends, weather-related losses, legal reforms, court

decisions, the level of economic activity, and other macroeconomic factors. It also includes the risk to smaller and specialty lines of business caused by a lack of credible data. For many lines of business systemic risk is the major component of underwriting volatility. The systemic risk factors for major lines by region appear on the next page. Detailed charts comparing motor and property risk by country appear below. The factors measure the volatility of gross loss ratios. If gross loss ratios are not available the net loss ratio is used.

Coefficient of Variation of Gross Loss Ratio by Country Property

Motor Israel Japan France Taiwan South Korea Hungary Australia Switzerland Austria Spain Czech Republic Germany Mexico Bolivia Turkey Uruguay Italy India Netherlands Chile Brazil Pakistan Colombia Dominican Republic Poland Canada Argentina Malaysia Vietnam South Africa U.K. U.S. Peru El Salvador Singapore China Honduras Venezuela Slovakia Indonesia Denmark Ecuador Panama Romania Hong Kong Nicaragua Greece Philippines

5% 5% 6% 7% 7% 8% 8% 8% 8% 8% 9% 9% 9% 10% 11% 11% 12% 12% 12% 12% 13% 13% 13% 13% 14% 14% 14% 15% 15% 15% 15% 16% 16% 17% 18% 18% 18% 18% 22% 22% 22% 23% 25% 37% 41% 46% 50%

Americas

4

Denmark Israel South Africa Australia Germany Spain Austria Italy Switzerland U.K. Canada Bolivia Chile Netherlands Turkey Malaysia Japan France China Poland India Hungary Uruguay Venezuela El Salvador South Korea Vietnam Ecuador U.S. Honduras Panama Argentina Pakistan Colombia Slovakia Nicaragua Romania Dominican Republic Hong Kong Brazil Indonesia Singapore Philippines Greece Taiwan Mexico 64% Peru Asia Pacific

12% 14% 15% 16% 16% 17% 17% 18% 20% 21% 22% 23% 25% 25% 26% 27% 28% 28% 30% 33% 34% 34% 35% 37% 38% 42% 42% 43% 43% 45% 46% 47% 51% 53% 53% 55% 57% 61% 62% 65% 66% 68% 70% 77% 84% 93% 96% Europe, Middle East & Africa

Aon Benfield

Fidelity & Surety

Credit

29%

16%

Workers’ Compensation

Marine, Aviation & Transit

57%

23%

General Liability

47%

10%

Property

14%

Bolivia

Motor Personal

Argentina

Motor

Accident & Health

Property Commercial

Property Personal

Motor Commercial

Underwriting Volatility for Major Lines by Country, Coefficient of Variation of Loss Ratio for Each Line

Americas 9%

274%

Brazil

13%

65%

50%

72%

70%

49%

45%

72%

Canada

14%

22%

17%

36%

35%

42%

78%

96%

Chile

12%

25%

49%

58%

23%

Colombia

13%

53%

25%

16%

76%

68%

104% 52%

Dominican Republic

13%

61%

86%

72%

Ecuador

23%

43%

51%

92%

El Salvador

17%

38%

15%

97%

Honduras

18%

45%

10%

243%

Mexico

9%

93%

Nicaragua

46%

55%

58%

Panama

25%

46%

18%

Peru

16%

96%

Uruguay

11%

U.S.

16%

Venezuela

18%

63%

65%

44% 170% 99%

23%

23%

53%

39%

130%

35% 14%

24%

43%

48%

34%

38%

37%

27%

69%

20%

230%

Asia Pacific Australia

8%

16%

23%

32%

54%

30%

63%

38%

29%

20%

13%

86%

22%

65%

13%

30%

66%

132%

52%

65%

5%

28%

12%

10%

16%

6%

Malaysia

15%

27%

119%

30%

36%

89%

Pakistan

13%

51%

Philippines

64%

70%

Singapore

18%

68%

China

18%

Hong Kong

41%

India

12%

Indonesia

22%

Japan

16%

62% 34% 29%

10%

127% 78% 87%

123%

36% 72%

74%

87%

41%

75%

South Korea

7%

7%

42%

33%

Taiwan

7%

7%

84%

50%

23%

73%

15%

42%

38%

11%

30%

Austria

8%

17%

13%

21%

13%

20%

Czech Republic

9% 22%

12%

15%

15%

18%

18%

27%

6%

28%

31%

23%

29%

25%

48%

17%

30%

26%

19%

22%

83%

84%

Vietnam

30%

163% 40%

52% 79%

Europe, Middle East & Africa

Denmark France Germany

49%

50% 25%

9%

16%

50%

77%

Hungary

8%

34%

Israel

5%

14%

56%

Italy

12%

18%

25%

19%

43%

42%

Netherlands

12%

25%

25%

51%

34%

33%

67%

35%

39%

33%

13%

33%

46%

22%

11%

50%

81%

26%

38%

12%

61%

80%

27%

33%

5%

Greece

Poland

14%

33%

Romania

37%

57%

Slovakia

22%

53%

South Africa

15%

Spain

8%

Switzerland

8%

Turkey

11%

U.K.

15%

17%

12%

21%

20% 15%

8%

26%

19%

21%

64%

88%

15% 8%

50%

20%

112%

Reported CVs are of gross loss ratios, except for Argentina, Australia, Bolivia, Chile, Ecuador, India, Malaysia, Singapore, Uruguay, and Venezuela, which are of net loss ratios. Accident & Health is defined differently in each country; it may include pure accident A&H coverage, credit A&H, and individual or group A&H. In the U.S., A&H makes up about 80 percent of the “Other” line of business with the balance of the line being primarily credit insurance.

5


U.S. Risk Parameters The U.S. portion of the Insurance Risk Study uses data from ten years of NAIC annual statements for 2,319 individual groups and companies. The database covers all 22 Schedule P lines of business and contains 1.7 million records of individual company observations from accident years 1992–2011. The charts below show the loss ratio volatility for each Schedule P line, with and without the effect of the underwriting cycle. The effect of the underwriting cycle is removed by normalizing loss ratios by accident year prior to computing volatility. This adjustment decomposes loss ratio volatility into its loss and premium components. Coefficient of Variation of Gross Loss Ratio | 1992–2011 All Risk Private Passenger Auto Auto Physical Damage Commercial Auto Workers’ Compensation Warranty

No Underwriting Cycle Risk 14% 17%

13%

Auto Physical Damage

15%

24% 27%

Commercial Auto

18%


19%

31%

Commercial Multi Peril

34%

Medical PL – Occurrence

35%

Other Liability – Occurrence

Private Passenger Auto

Warranty Commercial Multi Peril Medical PL – Occurrence

38%

Other Liability – Occurrence

Special Liability

39%

Special Liability

Other Liability – Claims-Made

41%

Other Liability – Claims-Made

Medical PL – Claims-Made

42%

Medical PL – Claims-Made

Products Liability – Occurrence

47%

Products Liability - Occurrence

Homeowners

48%

Homeowners

Other Reinsurance – Liability Fidelity and Surety International Reinsurance – Property Reinsurance – Financial Special Property Products Liability – Claims-Made

36%

27% 30% 27% 31% 34% 41%

Other

53%

50%

Reinsurance – Liability

67%

44%

Fidelity and Surety

69% 85% 93%

56%

Reinsurance – Property

57%

Reinsurance – Financial

61%

Special Property

58%

98%

Products Liability – Claims-Made

100%

U.S. Risk Parameters Over Time The Insurance Risk Study, now in its seventh edition, has published gross loss ratio CVs for Schedule P lines of business each year since its inception. The first edition CVs were based on data from 1992 — 2005. Since then, six additional years of results have added 43 percent more data to the analysis. The table on the right compares the published CVs from the first edition of the Insurance Risk Study to the current CVs. For many lines of business the CV is remarkably consistent from the first edition to seventh edition, showing that the methodology for estimation of CV is robust. Those lines which have experienced large changes in estimated CV from the original study are mostly longer tailed casualty lines that have seen significant loss development on old accident years and greatly improved pricing in recent years.

51%

International

72%

Financial Guaranty

143%

49%

Financial Guaranty

100%

Coefficient of Variation of Gross Loss Ratio Line of Business

1st Edition

7th Edition

Change

Private Passenger Auto

14%

14%

0%

Commercial Auto

24%

24%

0%

Workers' Compensation

26%

27%

1%


32%

34%

2%

Medical Malpractice — Claims-Made

33%

42%

9%

Medical Malpractice — Occurrence

35%

35%

0%

Other Liability — Occurrence

36%

38%

2%

Special Liability

39%

39%

0%

Other Liability — Claims-Made

39%

41%

2%

Reinsurance — Liability

42%

67%

25%

Products Liability — Occurrence

43%

47%

4%

International

45%

72%

27%

Homeowners

47%

48%

1%

Reinsurance — Property

65%

85%

20%

81%

93%

12%

102%

100%

-2%

Reinsurance — Financial Products Liability — Claims-Made

6

32%

28%

Aon Benfield

U.S. Reserve Adequacy and Risk Prognostications about the adequacy of U.S. statutory reserves are a favorite news item, with many different organizations entering opinions each year. The prognosticators are helped by high quality, uniform data at the legal entity level available through the NAIC Schedule P in statutory accounts. These accounts provide U.S. regulators with a clear view into insurance companies and are part of a very effective system of solvency regulation based on consistent and transparent reporting, albeit a system operating in a different way than the proposed Europe Solvency II protocol.

We estimate that companies will release USD7-10 billion of favorable development in 2012, and that the redundancy will be eliminated in one to two years. Through six months of 2012 companies have actually released USD8.2 billion of reserves, compared to USD7.8 billion in 2011. The reserve redundancy in the two most recent accident years has decreased from USD11 billion to USD3 billion, reflecting continued market pricing pressures. At year end 2011, accident years 2010 and 2011 accounted for 45 percent of the total industry booked reserves. As we have discussed in prior Insurance Risk Studies, understanding reserve risk is critical for effectively modeling company solvency. It is also a notoriously difficult problem: whereas all companies face broadly similar insurance risks, such as weather, legal, social and medical trends, each company’s reserving practices are idiosyncratic. There have been calls by some leaders in the industry for an approach based more on macro drivers: understand aggregate industry losses and apply market share to determine an initial estimate, or, possibly a prior estimate to weight with company specific estimates. The success of simple projections from Schedule P data indicate the macro drivers approach would increase the robustness and reliability of industry reserve estimates. Aon Benfield Analytics has developed effective models of industry loss drivers for some U.S. lines and continues to work to expand its understanding of macro drivers across all classes of business.

Three years ago, Aon Benfield started publicly tracking the reported reserve adequacy of U.S. companies. Each year we analyze the aggregated net loss development data by Schedule P line of business. Working at an aggregate level allows our actuaries to use different methods, and to weight the results in different ways than is possible for company actuaries who are working with smaller and less stable data sets. Unlike many other public studies, each of our reports has called for continued reserve releases by the industry — predictions that have been borne out by subsequent facts. The table below summarizes our analysis of the year end 2011 data. Our analysis indicates an overall redundancy of USD11.7 billion or 2 percent of total booked reserves, compared to USD22.0 billion at year end 2010. The bulk of the redundancy is in personal lines (USD7.6 billion) and commercial liability, including general liability, professional and medical malpractice (USD6.7 billion). Commercial property is very slightly redundant and workers’ compensation has a USD1.7 billion deficient.

U.S. Reserve Estimated Adequacy and Development Summary (USD Billions) Booked Reserves

Remaining Redundancy at YE 2011

2008

132.5

140.1

7.6

44.1

45.1

Commercial Liability

228.8

Workers’ Compensation Total Ex. Financial Guaranty

Line Personal Lines Commercial Property

Financial Guaranty Total

Estimated Reserves

Favorable / (Adverse) Development Average

Years at Run Rate

2009

2010

2011

5.4

5.8

6.7

3.4

5.3

1.4

1.0

2.6

2.4

2.7

2.6

2.6

0.4

235.5

6.7

5.2

3.8

2.4

7.2

4.6

1.4

140.7

138.9

(1.7)

1.1

(0.5)

(1.6)

(0.0)

(0.2)

N/A

546.1

559.6

13.5

14.4

11.5

10.1

13.1

12.3

1.1

27.3

25.4

(1.8)

(12.6)

7.0

0.4

(0.4)

(1.4)

N/A

573.4

585.1

11.7

1.7

18.6

10.5

12.7

10.9

1.1

7


Modeling the Underwriting Cycle The underwriting cycle is an important and much discussed insurance phenomenon. It can be seen most clearly in the ratio premium to GDP (shown right as the gold line). Premium to GDP does a good job of focusing on fundamentals. In other views of the cycle, such as combined ratio or ROE, the essential dynamics tend to be obscured by catastrophe losses and other artifacts. We are now eight years into the current softening phase that began in 2004. What causes the cycle, and what do the causes suggest for the chance it will turn over in the coming years?

Industry NWP to GDP % 4.5%

Wheat

Actual Fitted

4.0%

3.5%

3.0%

2.5%

The brown line in the chart is a simple model fit to the cycle that explains 90 percent of the variability in the data since 1970. The model fits this year’s premium to GDP ratio based on three variables: the prior year ratio, the second prior year ratio and the prior year ratio of loss to GDP. The parameters and standard errors from the model are shown in the table. The explanatory variables for this autoregressive model fall into two categories. The loss component is the signal in the model providing the fundamental driving force. The two premium components are characteristic of a dampedoscillation noise around the signal. They capture the well-known lags in pricing and reserving that tend to produce an over-reaction to a new loss shock taking many years to work through the system. Reassuringly, the model fits both the data and provides a behavioral explanation of the cycle consistent with the conventional wisdom! What does the model say about the potential for a market turn? The model says clearly that the turn will be loss driven. None of the market turns in the last forty years occurred without a major loss trigger. To explore the question more quantitatively, we fit the linear model using Bayesian Markov Chain Monte Carlo methods. The parameters are almost identical to those above, but the model is now fully stochastic and allows us to project ranges for the premium to GDP ratio in 2012 and beyond. The full model uses an auto-regressive process for GDP and splits losses into non-catastrophe

8

1965

1970

1975

1980

1985

1990

1995

2000

2005

2010

Cycle Model Statistics Parameter

Std. Error

Constant

0.0106

0.0018

Prior Loss/GDP

0.4512

0.0870

Prior Premium/GDP

1.0156

0.1187

Second Prior Premium/GDP

-0.6706

0.0983

Regression R 2

89.9%

(based on a normal distribution with a CV of 15 percent) and catastrophe components (based on a detailed analysis of net losses built up by company). The results show nominal net written premium almost certainly increasing in 2012 — consistent with results at six months — and 84 percent certain to increase in 2013. By 2014 there is a 63 percent chance of an increase but 37 percent chance of a decrease. The model indicates an industry loss USD50 billion above normal in 2012 would produce a 12 percent increase in premium in 2013; while a loss of USD150 billion above normal produces a hard market with a 22 percent increase.

Aon Benfield

New Paradigm Optimization

The standard optimization framework consists of a portfolio of potential risks of known profitability, which can be written in whole (an integer programming problem) or in part. In addition, there is a risk constraint that links portfolio risk to the amount of capital required. Risk constraints are typically expressed as one or more value at risk or tail value at risk constraints. They may come from regulator or rating agency models, in which case return periods are explicit, or from more general economic considerations, when return periods are more loosely defined. Instead of the standard framework, we can re-phrase the optimization problem in order to remove arbitrary choices and to get a more informative answer. A productive approach is to ask for the largest portfolio that covers its cost of capital. A simple implementation will have a single cost of capital, and the optimization becomes “maximize premium volume subject to the requirement that expected profits are greater than the cost of capital times capital required for the portfolio”. Cost of capital is a market derived input and typically the capital requirement is externally specified. The optimization has no free variables to be selected. The amount of risk in the optimal portfolio is a result of the optimization as opposed to an arbitrary input. And the output portfolio is guaranteed to be an attractive business prospect: by specification it meets its own cost of capital. There is one downside: the optimal solution may be to write no business — optimization can only help a portfolio so much. We call this approach “New Paradigm Optimization” or NPO.

Risk Return Optimization

Reward (Premium)

Techniques to optimize a portfolio to achieve the lowest risk for a given return (or highest return for a given risk) are a standard part of the analytics toolkit. Aon Benfield Analytics has developed several optimization tools, each tailored to a different client problem. They range from Dynamic Portfolio Optimization (DPO) to customized programming, and include portfolio pricing techniques such as Cat Score® that also provide underwriting insights. Two challenges with many existing methods are the lack of guidance they offer about the acceptable amount of risk and their failure to guarantee that an optimized portfolio is actually an attractive business prospect. New techniques are now available to address both of these challenges.

?

Efficient Frontier (with constraints)

Current Portfolio

Risk A more sophisticated NPO allows for a full capital tranching, with different costs of capital for equity, lower and higher rated debt. The model can be set up to include real-world constraints limiting the ratio of debt to equity and the relative amounts of high and low credit quality debt. Again, all the required debt spread and cost of equity inputs are available in the market, and rating agency leverage constraints are externally specified. Another variant of NPO considers the cost of equity capital to be an input variable. The model can then be run as a function of cost of capital, producing different portfolios for each target return. Higher returns produce smaller portfolios because fewer risks will meet the more stringent profit requirement. The schedule of premiums versus returns can then be combined with a valuation model, such as Aon Benfield’s price to book regression study, in order to maximize valuation. The regression study fits price to book valuation as a function of prospective ROE (i.e., cost of capital). The combined model finds the portfolio — and attendant level of risk — to maximize corporate value. The only inputs to the model are the universe of potential risks, the capital requirement formula and the valuation formula. The optimal portfolio is completely specified by these inputs. Aon Benfield Analytics’ portfolio optimization team is available to explore application of its broad toolkit to your particular optimization problem. The new approaches described above have already been used effectively in personal lines and reinsurance settings. To find out more, contact your Aon Benfield broker or one of the contacts listed on the last page.

9


Motor Insurance in China Motor insurance is the largest line of business in China, constituting over 70 percent of premium in the property and casualty market. Since the product’s inception in China it has been a heavily regulated line of business. Authorization to write business, pricing, product design, and policy wordings have historically been under the tight control of the China Insurance Regulatory Commission (CIRC). This has severely limited product differentiation and innovation amongst Chinese insurers. But as with everything else in China, regulation is rapidly changing.

80 70

250

60

200

50

150

40 30

100

20

50

10

0 2001

2002

2003

2004

* Motor Premium Estimated RMB350B = USD55B = EUR44B

10

2005

2006

2007

2008

2009

2010 2011*

0

assistance for policyholders

and sales online as well as using social media for brand development

Vehicle Ownership (Millions)

Motor Premium (RMB Billions)

90

300

• Bundling of services such as unlimited free roadside

• Expansion into the internet space, providing quotes

China Motor Premium

350

commission sale of motor insurance

For example, one insurer promises to settle all claims smaller than RMB10,000 (about USD1,600) within one working day

Since 2001, China’s motor insurance market has grown by more than 20 percent each year. This enormous growth has been driven by a five-fold increase in Chinese auto ownership over the same time period. Although current economic conditions and changes in regulation for auto ownership have the potential to dampen growth in the short term, a steady increase in motor insurance premiums is expected in the long run.

100

• Development of telemarketing platforms for low

• Simplification of claim-handling processes for customers.

In recent years, the CIRC has implemented regulatory reforms and has signaled future reforms that will allow greater competition in the marketplace and the entry of foreign insurers. These changes offer early adopters the opportunity to materially increase market share in a low volatility line of business. Experience in the U.S., Australia and Europe shows that first movers are able to write profitable risks while those that are slow to adapt suffer from adverse selection.

400

The Chinese motor insurance marketplace has witnessed significant innovation as the CIRC has relaxed other rules and regulations around the product. Some of the key developments by insurers in the recent years are:

Starting in 2013, the CIRC will allow insurers meeting certain criteria to develop and set rates based on their own underwriting data and loss experience. As a result, pricing sophistication is expected to materially increase as companies develop their own predictive models or hire consultants with pricing expertise to help create their rating plans. This change is also expected to generate strong price competition and promises to alter the motor insurance landscape the same way that predictive modeling has changed motor insurance in other markets around the world. The deregulation of China motor insurance to allow entry from foreign insurers poses a once in a lifetime opportunity. Prepared companies have a chance to establish market share in a rapidly growing marketplace. Data systems, pricing techniques, product innovation, and distribution channels will be among the top priorities for many Chinese insurers in the coming years. As a global organization that has been involved in similar reforms in other parts of the world, Aon Benfield is committed to helping our clients reach a strategically established position by bringing our experience and expertise to the Chinese auto insurance market.

Aon Benfield

A Different View of Risk The following pages present different views of the world — views based on population, economics and exposure to risk. A series of Aon Benfield cartograms highlight global insured risk concentrations and largely explain relative international catastrophe reinsurance pricing. The cartogram maps and underlying data can be used as a traditional backup to model-based catastrophe risk management. Map 1 is called a cartogram. In a cartogram the area of each country is scaled to be proportional to a social or economic variable. In Map 1 the area of each country is proportional to its population. The cartogram is created using Gastner and Newman’s algorithm which preserves actual geographical shapes as closely as possible. Map 1: Population

The 12 countries with populations over 100 million (in order: China, India, United States, Indonesia, Brazil, Pakistan, Nigeria, Bangladesh, Russia, Japan, Mexico, and Philippines) stand out very clearly. The heavily populated Eastern seaboard of the United States, the Low Countries of Netherlands, Belgium and Luxembourg, and Eastern United Kingdom are also clearly visible. Populations within country are based on estimated populations per square mile, rather than assuming a constant average population over the entire

country. The illustration provides insight into likely human casualties from a catastrophic event. In the map countries are colored by continent, with Europe split into western and eastern, and Africa split into northern and southern groups. These colors will be used in subsequent maps for reference. The next pages will show a progression of cartograms to illustrate global human, economic and insured risk relativities.

11


Map 2 shows the globe using the Gall-Peters projection. The area of each country on the map is proportional to its actual land area. Compared to more common projections the Gall-Peters appears to shrink various northern hemisphere countries (Greenland) that are often grossly distorted relative to more populous equatorial and southern hemisphere countries. In Map 3 the area of each country is proportional to its latest reported GDP on purchasing parity basis (data sources for all the maps are given on page 34). The fifteen largest countries are now, in order: United States, China, Japan, India, Germany, United Kingdom, France, Russia, Brazil, Italy, Mexico, Spain, Canada, South Korea, and Indonesia. Again, various smaller countries in Western Europe are clearly visible, whereas poorer countries in Africa and Asia are considerably smaller. GDP within country is spread according to population. Map 3 can be used to gauge the potential for economic loss from a catastrophic event.

12

Map 2: Gall-Peters Projection

Map 3: GDP

Aon Benfield

Map 4 provides another view of economic exposure by making the area of each country proportional to its estimated wealth. Countries like China and India have experienced very substantial GDP growth over the last decade but lack the depth of infrastructure investment accumulated over many decades in the U.S. and Map 4: Wealth Europe, and so appear relatively smaller than in Map 3. The wealth estimates use the latest World Bank Wealth of Nations report. Wealth includes natural capital (renewable and nonrenewable resources), produced capital (machinery, equipment and structures) and intangible (human) capital. Liability insurance exposure to loss of life varies substantially between countries depending on accumulated human capital. The fifteen wealthiest countries are estimated to be

(in descending order): United States, Japan, Germany, United Kingdom, France, Italy, China, Spain, Canada, Brazil, Mexico, South Korea, India, Australia, and Russia. The ratio of wealth to GDP varies from less than 2 (China) to the mid-teens (U.S., Japan, Western Europe).

Map 5 illustrates property and casualty insurance premium and insurance penetration. Area is proportional to insurance penetration, defined as premium to GDP ratio times wealth. If all Map 5: Insurance Penetration countries were similarly insured it would be identical to wealth, Map 4. But a high wealth country with low insurance penetration (e.g., India) drives less insurance exposure than a similarly wealthy country with high penetration (e.g., Australia). Map 5 accounts for these differences in insurance penetration. It highlights the geographic concentration of the countries served by the insurance and reinsurance industries. Insurance effectively serves three areas: North America, Western Europe, and

Japan and South Korea. These regions account for USD1.00 trillion out of USD1.28 trillion global premium (80 percent), but only 15 percent of global population.

13


The main patterns of global earthquake incidence are well known: the Pacific “ring of fire”, which accounts for over 80 percent of the world’s largest quakes, Central Asia and Eastern Europe. On a GDP basis (Map 7) the disproportionate hazard from California and Japan become very clear. Maps 6-9 show global earthquake damage potential from ground shake in a 50-year time window. The maps are based on Global Seismic Hazard Assessment Program data. The GDP transformation collapses the size of faults running through India, Nepal, Pakistan, and Afghanistan but somewhat expands them in the Middle East and Eastern Europe, reflecting relative levels of GDP and economic exposure. The New Madrid seismic zone in the Central U.S. appears relatively small on the maps because its last major event was in 1811-12, over 200 years ago. Risk management for the New Madrid fault requires a catastrophe model because of its extremely low frequency.

14

Map 6: Earthquake Gall-Peters Projection

Map 7: Earthquake GDP Transformation

Aon Benfield

Detailed maps of Japan and California show how population centers align closely with the highest hazard areas, especially around Tokyo and Los Angeles.

Map 8: California Earthquake Gall-Peters Projection

Population Transformation

Map 9: Japan Earthquake Gall-Peters Projection

Population Transformation

15


Risk from various forms of severe wind affects many coastal and inland areas. The Aon Benfield cartogram (Map 11) shows very high risk concentrations along the eastern coasts of the U.S., China and Taiwan. Somewhat lower hazard winds affect very substantial exposures in the U.S. Midwest and Eastern Europe. Maps 10 and 11 combine and smooth up to one hundred years of historical data for hurricanes, typhoons, tornados and severe convective storms (U.S.), and winter storms (Europe). Impact Forecasting combined a variety of data sources, including the International Best Track Archive for Climate Stewardship, to produce the maps. The representative size of a typical event leads to differences in potential occurrence and aggregate losses. Typhoon and hurricanes are the most destructive events. In Europe large winter storm events cause substantial economic loss. Midwestern U.S. wind losses, while locally more extreme, are typically smaller and so result in lower industry occurrence losses but, as seen in 2011, higher aggregate losses.

16

Map 10: Wind Gall-Peters Projection

Map 11: Wind GDP Transformation

Aon Benfield

Low lying areas are among those susceptible to flood. Maps 12 and 13 highlight areas less than 10 meters above sea level as well as 60 recent serious flood events. Ten representative events are described below.

1

2008 U.S. Midwest Floods: 17 deaths, USD15 billion economic loss, USD8 billion insured loss

2

1993 U.S. Mississippi River Floods: 50 deaths, USD21 billion economic loss, USD2 billion insured loss

3

1999 Venezuela Floods & Landslides: 30,000 deaths, USD3.1 billion economic loss

4

2002 Central Europe Floods: 55 deaths, USD17 billion economic loss, USD3.5 billion insured loss

5

1998 Bangladesh & India Floods: 3,838 deaths, USD4.3 billion economic loss

6

2004 Indonesia EQ & Tsunami: 227,898 deaths, USD14 billion economic loss, USD5 billion insured loss

7

2011 Thailand Floods: 813 deaths, USD45 billion economic loss, USD15.5 billion insured loss

8

1998 China Floods: 3,656 deaths, USD30 billion economic loss, USD1 billion insured loss

9

2011 Japan EQ & Tsunami: 15,868 deaths, USD210 billion economic loss, USD35 billion insured loss

Map 12: Flood Gall-Peters Projection

Map 13: Flood GDP Transformation

10 2010-2011 Australia Floods: 36 deaths, USD30 billion economic impact, USD2.4 billion insured loss

17


Global Wind and Earthquake Risk Map 14: Wind and Earthquake GDP Transformation

Wind and earthquake are the two most severe insured perils and together they drive global insurance capital requirements. Exposures to both are combined on the main map with area proportional to GDP producing a good representation of global exposure to catastrophic economic loss. As expected, the majority of the map is exposed to one or both perils. Several areas stand out:

• The size and extreme hazard of hurricane exposure to the Eastern U.S., which in aggregate is the global peak economic exposure • The major earthquake exposures in Japan and California • Japan’s additional exposure to typhoon loss — indeed, most of the country is exposed to both perils • Taiwan’s economic exposure to wind and earthquake • Western Europe’s very large economic exposure to winterstorm

18

Aon Benfield

Map 15: Wind and Earthquake Insurance Penetration Transformation

Insured Loss Potential U.S. Hurricane Japan Earthquake U.S. Earthquake Europe All Perils Japan Typhoon Aus & NZ All Perils Other Perils

• Central and Eastern Europe’s exposure to earthquake • The substantial and quickly growing exposure to typhoon along China’s extensive East Coast Insurance penetration varies by line of business and by country, and the amount of insurance in the private market is affected by government pools. Map 15 shows combined wind and earthquake exposures with area proportional to insurance penetration. The relative importance of global perils to the reinsurance market is now much clearer, and is

shown in the bar chart. Japan earthquake is estimated gross of the Japanese Earthquake Reinsurance pool; on a net basis it is reduced by approximately 50 percent. Accounting for insurance penetration for China and Taiwan deliver less insured risk into the global market than their economic risk. China’s rapid development and increased insurance penetration will substantially increase its importance over time. The exposures behind the bar chart are consistent with global patterns in catastrophe reinsurance pricing.

19


Robust Risk Management Several catastrophic events during 2011 caused surprisingly large losses to the insurance industry, including the Tōhoku earthquake, Thailand floods and killer tornados in the U.S. These events underscore the importance of robust risk management to supplement sophisticated simulation modeling. Intelligent visualization of exposure, leveraging lessons from historical events and deterministic scenarios are all important parts of comprehensive and robust risk management. The Aon Benfield cartograms shown on the previous pages are an excellent way to understand concentrations in an insurance portfolio. Maps 8 and 9 show how, combined with risk hazard assessments, cartograms deliver greater insight into risk drivers than traditional maps which only display geographical area. Maps 12 and 13 show what can be learned from historical flood events. They help identify at-risk areas that should be prioritized for further risk management via value and insured limit concentration monitoring and reporting. Three ingredients are needed in order to use exposure information to derive robust “what-if” analyses for risk management. • How frequent are representative events? The maps and cartograms, and the data underlying them, provide a rich source of information on absolute and relative frequency for wind and earthquake. They can be used to prioritize additional risk management efforts globally. • How large are representative events? For a hurricane or typhoon, what is the radius of damaging wind speeds? For a tornado, what is the path length and width? For flood, understanding topography (Map 13), bathymetry and river valley geography will inform potential event sizes. • Within each event, what range of damage can be expected? Newsreels show sensational pictures of very local total devastation caused by catastrophic events, but they rarely zoom-out to show the bigger picture. Within a band of hurricane or typhoon winds, what is a typical damage ratio? What about within a tornado track or the path of a broad winter storm? A good example where we have answers to these three questions and can effectively supplement catastrophe models is for U.S. tornadoes. Aon Benfield Analytics has used this approach to produce Tornado Viewing Guides for clients. They compute what-if losses from three major historical tornadoes — only allowing them to touch-down anywhere east of the Rockies. The potential frequency of extreme tornadoes

20

Joplin, MO Tornado Date:

May 22, 2011

Rating:

EF-5

Est. Max. Wind:

200+ mph

Path Length:

22.1 Miles

Path Width:

0.75 Mile

Fatalities

162

can be estimated from the historical record. It shows about one F4 or F5 tornado per 500,000 square kilometers per year. The average frequency can be further refined by state. Maps 10 and 11 illustrate approximate frequency for the Midwest. The path and pertinent information for 2011’s Joplin tornado is shown above. It is a good representative example of a severe tornado event. We can estimate the average loss to all properties in the track by analyzing losses from large portfolios exposed to Joplin and other major tornadoes. Losses typically fall in the range 30 to 80 percent of values. Further adjustments are then needed to estimate total losses from the super-cell outbreak as opposed to the individual track loss. Tornado Viewing Guides have proven very helpful to management as they struggle to understand the potential for extreme losses in their portfolios in a tangible and credible manner. The output can easily be explained to Boards or risk management committees, and its basis in exposure information provides a natural way to manage, monitor and limit exposure. Finally, the assessment of frequency is helpful to place tornado risk in context with other major risks the organization faces.

Aon Benfield

Global Crop Insurance Penetration In last year’s Insurance Risk Study, we discussed Aon Benfield’s Crop Reinsurance Solutions (ACReS) models for the U.S. and China. Both markets have seen substantial increases in crop insurance premium over the last 10 years and China, in particular, has potential for continued significant growth. In 2011 alone, China’s crop insurance premium increased by 28 percent. Crop insurance continues to expand globally and is currently available in some form in 105 countries around the world. Significant expansion efforts are being made in China, India, Pakistan, Brazil, Ukraine, Columbia, Vietnam, South Korea, Malaysia, and Thailand. International aid projects encouraging the development of agricultural risk management programs are also in progress in Sub-Saharan Africa, Ukraine, India, and Bangladesh. Despite these efforts, many regions with high agricultural production continue to have low levels of market penetration. For example, only 25 million of 120 million farms, or 22 percent of the cropped area, is insured in India. Brazil has relatively low participation with about 33 percent

of the cropped area insured. In China, less than 15 percent of the total value of crop production is insured. Clearly there is room for expansion in these and many other countries around the world. In the map below, regions are shaded based on agricultural production. In addition, each region is labeled as high, medium or low insurance market penetration. The U.S. is the only large agricultural producer with high market penetration. China, on the other hand, is considered to have low crop insurance penetration rates despite the largest worldwide crop exposure. India, Brazil, Argentina, Russia, and Ukraine also have high crop exposure with low insurance penetration. As the world’s population grows, the demand for agricultural outputs will increase. As crop prices and volatility rise, the world will have a larger need for systems to effectively manage the risk associated with varying levels of crop prices and production. The map clearly illustrates the fact that there are many opportunities for the insurance industry to expand and improve their crop insurance offerings globally.

Agriculture Production (Million Metric Tons) 0 to 50 50 to 100

Agriculture Production (Million Metric Tons) 0 to 50 50 to 100 100 to 200 200 to 400 over 400

100 to 200 200 to 400 over 400 Crop Insurance Penetration High Medium Low

Crop Insurance Penetration High Medium Low

21


Correlation Considerations Correlation between lines of business is central to a realistic assessment of aggregate portfolio risk, and in fact becomes increasingly significant for larger companies where there is little idiosyncratic risk to mask correlation. Most modeling exercises are carried out at the product or business unit level and then aggregated to the company level. In many applications, the results are more sensitive to the correlation and dependency assumptions made when aggregating results than to all the detailed assumptions made at the business unit level. The Study determines correlations between lines within each country. Correlation between lines is computed by examining the results from larger companies that write pairs of lines in the same country. The table below shows a sampling of the results available for the U.S. The matrix is organized along major line of business categories: personal lines, standard commercial lines and specialty commercial lines. It shows that correlation between personal lines is smaller than the correlation between standard commercial lines. Specialty commercial lines tend to exhibit the strongest between line correlations.

However, correlation varies significantly with premium volume. The U.S. correlation coefficients below represent an average level of correlation for companies with premium volume above a threshold of USD100 million. We selected this threshold as representative of the size of a typical product division within a medium to large insurance company. The observed level of correlation varies with this threshold. Companies with volume exceeding USD100 million will observe an increasing level of correlation between lines. For example, between workers’ compensation and other liability occurrence, the correlation above a USD100 million threshold is 61 percent, above USD500 million it is 73 percent and above USD1 billion it is 85 percent. Correlation increases as the premium threshold increases across most lines of business. The larger a company becomes, the more important correlation will be for the company. Regulators and rating agencies scrutinize correlation assumptions in their evaluations of capital modeling practices. Aon Benfield Analytics is ready to help companies understand the implications of their correlation assumptions, design custom analysis to facilitate parameterization of correlation assumptions and guide companies through the rating agency review process.


Commercial Auto


Other Liability Occ

Medical Malpractice CM

Other Liability CM

Products Liability Occ

Homeowners

Personal Auto Liability

Homeowners

U.S.

8%

24%

9%

-2%

0%

2%

-2%

13%

30%

30%

33%

32%

51%

43%

45%

53%

43%

49%

57%

43%

40%

60%

67%

71%

44%

71%

61%

68%

62%

63%

77%

58%

66%

70%

72%

Personal Auto Liability

8%


24%

30%

Commercial Auto

9%

30%

53%


-2%

33%

43%

60%

Other Liability Occ

0%

32%

49%

67%

61%

Medical Malpractice CM

2%

51%

57%

71%

68%

77%

Other Liability CM

-2%

43%

43%

44%

62%

58%

70%

Products Liability Occ

13%

45%

40%

71%

63%

66%

72%

22

33% 33%

Correlation is a measure of association between two random quantities. It varies between -1 and +1, with +1 indicating a perfect increasing linear relationship and -1 a perfect decreasing relationship. The closer the coefficient is to either +1 or -1 the stronger the linear association between the two variables. A value of 0 indicates no linear relationship whatsoever. All correlations in the Study are estimated using the Pearson sample correlation coefficient. In each table the correlations shown in bold are statistically different from zero at the 90% confidence level.

Aon Benfield

Global Correlation between Lines The tables below show correlation coefficients between lines of business for China, Japan and U.K. Aon Benfield Analytics is capable of producing custom analyses of correlation for many insurance markets globally.

Credit

Engineering

General Liability


Motor

Property

Accident & Health

Agriculture

Accident & Health

China

32%

19%

17%

23%

16%

32%

55%

39%

26%

34%

20%

8%

3%

15%

16%

30%

21%

29%

14%

48%

28%

32%

37%

34%

Agriculture

32%

Credit

19%

39%

Engineering

17%

26%

39%

General Liability

23%

34%

15%


16%

20%

16%

14%

32%

Motor

32%

8%

30%

48%

37%

31%

Property

55%

3%

21%

28%

34%

8%

39%

29%

31%

8% 38%

38%

2%

48%

32%

47%

-3%

1%

33%

26%

12%

33%

-2%

55%

39%

-3%

Motor

48%

1%

12%

Property

32%

33%

33%

55%


47%

26%

-2%

39%


Property

2%

Motor


20%


20%

Accident & Health General Liability

General Liability

Accident & Health

Japan

31% 31%

52%

-47%

0%

42%

38%

14%

51%

40%

36%

Commercial Motor

77%

44%

Commercial Property

52%

38%

44%

Financial Loss

-47%

14%

-13%

-19%

Household & Domestic Private Motor

Private Motor

Financial Loss

77% 44%

Commercial Lines Liability

44%

Household & Domestic

Commercial Property

36%

Accident & Health

Commercial Motor

Commercial Lines Liability

Accident & Health

U.K.

-13%

17%

71%

-19%

68%

33%

-1%

-25%

0%

51%

17%

68%

-1%

42%

40%

71%

33%

-25%

17% 17%

23


Macroeconomic Correlation Correlation among macroeconomic factors is a very important consideration in risk modeling. The interaction of inflation and GDP growth with loss ratios and investment returns has a profound effect on insurer financial health and stability. The following matrix shows correlation coefficients for various macroeconomic variables impacting an insurer’s balance sheet. The Consumer Price Index and Producer Price Index are highly correlated, but they do not show particularly strong correlation with other factors. This may be because inflation has been relatively tame for the last 25 years. GDP growth shows strong negative correlation with changes in unemployment. When GDP drops — or unemployment increases — credit spreads tend to increase, property values fall and the VIX increases. Treasury yields and corporate bond spreads are inversely correlated; financial market fears may push investors to flee corporates for the safety of Treasuries, causing corporate yields to rise and Treasury yields to fall.

The VIX is sensitive to fear and directionally has the appropriate signs: positive correlation with spreads and unemployment, negative correlation with GDP and stock returns. These coefficients represent only the beginning of an analysis of macroeconomic dependency. Lags may be appropriate among certain variables. For example, GDP and stock returns show the strongest correlation when stock returns lead GDP by two quarters, suggesting that stock prices adjust as soon as expectations for GDP change, consistent with the Efficient Market Hypothesis. It is also important to consider values that shift over time. In successive eight-quarter periods, stock returns and property returns showed zero or negative correlations until the recent financial crisis when correlations turned strongly positive. Subsequent to the financial crisis this correlation has once again tempered. This fact alone suggests that a simplistic view of correlation among macroeconomic factors will significantly underestimate material balance sheet risks.

GDP Growth

Unemployment Change

3-Month T-Bill Rate

1-3 Year Treasuries

AAA-AA 3-5 Year Spread

BBB 3-5 Year Spread

S&P 500 Returns

VIX

Property Returns

Inflation (CPI-U)

Inflation (PPI)

Inflation (CPI-U)

Macroeconomic Correlation

50%

-6%

-1%

34%

39%

-9%

-15%

-10%

-12%

5%

4%

-8%

28%

6%

-6%

-18%

-6%

-22%

11%

-65%

-2%

26%

-65%

-71%

5%

-43%

53%

5%

8%

49%

51%

-2%

46%

-31%

98%

-29%

-58%

-7%

-23%

21%

-33%

-59%

12%

-24%

18%

84%

-42%

62%

-63%

-32%

66%

-54%

-50%

12%

Inflation (PPI)

50%

GDP Growth

-6%

4%

Unemployment Change

-1%

-8%

-65%

3-Month T-Bill Rate

34%

28%

-2%

5%

1-3 Year Treasuries

39%

6%

26%

8%

98%

AAA-AA 3-5 Year Spread

-9%

-6%

-65%

49%

-29%

-33%

BBB 3-5 Year Spread

-15%

-18%

-71%

51%

-58%

-59%

84%

S&P 500 Returns

-10%

-6%

5%

-2%

-7%

12%

-42%

-32%

VIX

-12%

-22%

-43%

46%

-23%

-24%

62%

66%

-50%

5%

11%

53%

-31%

21%

18%

-63%

-54%

12%

Property Returns

24

-31% -31%

Aon Benfield

Global Market Review With global capital levels approaching USD3.4 trillion and significant amounts of spare capacity available, insurers are eager for new sources of premium growth. The global market review section, explained on pages 25 to 30, identifies sources of current premium, potential growth markets, and loss ratio out or under performers among the 50 largest insurance markets globally. The table below summarizes current premium volume, GDP and premium to GDP ratio, a common measure of insurance penetration. Global Premium by Product Line

Top 50 Markets by Gross Written Premium

Motor: USD572B Brazil

Country

Canada Rest of Americas China

U.S.

Japan

South Korea Rest of APAC France

Middle East & Africa

Germany

Rest of Europe

U.K. Rest of Euro Area

Property: USD409B Brazil

Canada Rest of Americas China Japan South Korea Rest of APAC

U.S. France Germany U.K. Middle East & Africa

Liability: USD310B

Rest of Euro Area Rest of Europe

Brazil Canada Rest of Americas China Japan

U.S.

South Korea

Rest of APAC France Middle East & Africa Rest of Europe

Germany U.K. Rest of Euro Area

Notes: All statistics are the latest available. “Motor” includes all motor insurance coverages. “Property” includes construction, engineering, marine, aviation, and transit insurance as well as property. “Liability” includes general liability, workers’ compensation, surety, bonds, credit, and miscellaneous coverages.

U.S. Japan Germany France U.K. China S. Korea Italy Canada Australia Spain Brazil Netherlands Russia Switzerland Norway Belgium Austria Mexico Argentina Sweden Denmark Poland India Venezuela Turkey South Africa Iran Ireland Portugal Finland Czech Republic Malaysia Israel Colombia U.A.E. Chile Thailand Greece Taiwan New Zealand Luxembourg Indonesia Ukraine Hong Kong Singapore Romania Puerto Rico Saudi Arabia Hungary Grand Total

P&C GWP GDP (USD Billions) (USD Billions)

Population (Millions)

Premium / GDP Ratio

GDP per Capita

480.7 81.4 69.8 66.7 62.1 57.5 53.9 40.0 32.2 31.2 30.4 25.4 14.4 13.6 12.5 10.4 10.3 9.6 9.1 8.8 8.8 8.0 7.7 7.6 7.3 6.9 6.8

15,094.0 5,869.5 3,577.0 2,776.3 2,417.6 7,298.1 1,116.2 2,198.7 1,736.9 1,488.2 1,493.5 2,492.9 840.4 1,850.4 636.1 483.7 513.4 419.2 1,154.8 447.6 538.2 333.2 513.8 1,676.1 315.8 778.1 408.1

313.8 127.4 81.3 65.6 63.0 1,343.2 48.9 61.3 34.3 22.0 47.0 205.7 16.7 138.1 7.7 4.7 10.4 8.2 115.0 42.2 9.1 5.5 38.4 1,205.1 28.0 79.7 48.8

3.2% 1.4% 2.0% 2.4% 2.6% 0.8% 4.8% 1.8% 1.9% 2.1% 2.0% 1.0% 1.7% 0.7% 2.0% 2.2% 2.0% 2.3% 0.8% 2.0% 1.6% 2.4% 1.5% 0.5% 2.3% 0.9% 1.7%

48,094 46,083 43,995 42,302 38,345 5,433 22,846 35,891 50,637 67,599 31,748 12,118 50,233 13,401 83,084 102,745 49,184 51,004 10,044 10,610 59,122 60,114 13,375 1,391 11,261 9,757 8,360

4.3 4.3 4.1 4.1 4.0 3.8 3.8 3.7 3.6 3.5 3.4 3.3 3.0 2.9 2.8 2.7 2.7 2.1 2.1 2.0 1.8 1.8 1.7 1,244.4

482.4 217.7 238.9 266.6 215.3 278.7 242.9 328.4 360.1 248.4 345.6 303.1 466.8 161.9 58.4 845.7 165.0 243.3 259.8 189.8 96.3 577.6 140.3 65,201.0

78.9 4.7 10.8 5.3 10.2 29.2 7.6 45.2 5.3 17.1 67.1 10.8 23.1 4.3 0.5 248.2 44.9 7.2 5.4 21.8 4.0 26.5 10.0 4,859.3

0.9% 2.0% 1.7% 1.5% 1.9% 1.4% 1.6% 1.1% 1.0% 1.4% 1.0% 1.1% 0.6% 1.8% 4.8% 0.3% 1.6% 0.9% 0.8% 1.1% 1.9% 0.3% 1.2% 1.9%

6,117 46,097 22,157 50,647 21,151 9,550 31,999 7,260 67,767 14,555 5,152 28,145 20,197 37,397 114,742 3,407 3,678 34,012 48,538 8,686 24,072 21,768 14,089 13,418

Note: Ranks are based on total P&C Written Premium

25


Growth Markets and Out or Under Performers The following pages contain analysis of premium growth and loss ratio performance by country across motor, property and liability lines of business. The quadrant plots on this page summarize the results of that analysis and identify countries as either low growth or high growth and as loss ratio out performers or under performers for each line of business. The figures in the top right quadrants show the gross written premium size, in USD millions, of each country. Growth is determined based on five year annualized premium growth. Countries with values greater than 7.5 percent are classified as high growth.

Motor

Brazil China Colombia Indonesia New Zealand Norway Russia Saudi Arabia Thailand U.A.E Ukraine Venezuela

Low Growth

Argentina Australia Chile India Iran Malaysia Morocco Poland Singapore South Africa Turkey

Under Performers

26

Liability

Out Performers

14,827 44,361 1,690 982 1,058 3,258 7,780 864 2,354 1,231 775 5,263

Brazil Colombia India Indonesia Iran Romania Russia Saudi Arabia Thailand Turkey U.A.E Ukraine Venezuela

Hong Kong Hungary Japan Malaysia Norway Puerto Rico Singapore Spain Switzerland Taiwan U.S. High Growth

Belgium Canada Finland France Germany Ireland Israel Italy Mexico Portugal Romania S. Korea Spain Sweden U.K.

Five countries appear as high growth outperformers in each of the lines of business analyzed: Brazil, Columbia, Indonesia, U.A.E., and Venezuela. An additional ten countries appear in the upper right quadrant in two of the three lines of business analyzed. In total, there are 23 countries which are high growth outperformers in at least one line of business.

Property

Out Performers

Austria Czech Republic Denmark Hong Kong Hungary Japan Luxembourg Netherlands Puerto Rico Switzerland Taiwan U.S.

Loss ratio performance is determined based on five year cumulative loss ratio. Each country’s loss ratio performance is compared against their income level peers, using a USD30,000 GDP per capita threshold. Countries with five year loss ratios lower than the average of their income peers are classified as out performers.

Low Growth

Out Performers

Australia Canada France Hong Kong Hungary Ireland Japan Netherlands Puerto Rico Romania Russia Saudi Arabia South Africa Spain Switzerland Taiwan Thailand U.K. Ukraine

8,750 1,202 2,023 1,429 731 482 5,881 795 841 2,235 1,227 1,261 1,023 High Growth

Austria Belgium Canada Czech Republic Denmark Finland France Germany Ireland Israel Italy Netherlands Portugal S. Korea South Africa Sweden U.K.

Argentina Australia Chile China Luxembourg Mexico Morocco New Zealand Poland

Under Performers

Austria Brazil Chile China Colombia Czech Republic India Indonesia Iran Malaysia Mexico New Zealand Norway Poland Singapore Turkey U.A.E Venezuela

Low Growth

2,255 1,819 756 3,389 835 781 1,570 329 514 337 1,434 295 2,878 1,090 576 326 1,097 1,055 High Growth

Belgium Denmark Finland Germany Israel Italy Portugal Sweden U.S.

Argentina Luxembourg Morocco S. Korea

Under Performers

Aon Benfield

Target Options Size of opportunity

Distribution channel dynamics

By geography, LoB, & channel: • Available premium • Growth rates

• Broker & office preferences • Chanel organization • Client decision making

Insurer appetite & capabilities

Competitive behavior • Market share • Capacity • Terms & conditions • Appetite

Granular Growth Strategy In an increasingly complex global marketplace, insurers are seeking opportunities to write more profitable business. Inpoint, Aon Benfield’s consulting business, offers a broad set of service offerings leveraging unique access to placement data and granular market intelligence. We help insurers to prioritize among market opportunities and create business plans that are realistic, actionable, fact-based and, above all, profitable. Our proven strategic decision framework identifies accessible markets and high potential customer segments to formulate growth programs tailored to an insurer’s growth goals, capabilities and risk appetite. Inpoint works consultatively with insurers to systematically assess the key variables defining market attractiveness and profitable growth. The approach has four components. 1. Sizing the opportunity: Analysis of Aon’s proprietary data and market intelligence determines the target market size, assesses the attractiveness of various segments and identifies promising growth opportunities. Our data includes granular splits not available through public data, including size of market at the geographic, customer segment and line of business level. 2. Identifying distribution channel dynamics: Formulating the right channel strategy by line of business and customer type is critical to success. Selecting the proper channel and identifying what it takes for successful entry requires a deep understanding of key insurance buying factors and how business flows across geographies and placement markets. Inpoint helps its clients understand which characteristics are critical for success and how the leading insurers grow and defend their market share.

• Approach to broker channel • Risk appetite & capacity analysis • Target growth areas • Existing book characteristics

3. Assessing competitive behavior: Benchmarking against peers and leading competitors can identify performance gaps, provide insight into operating models and assist in setting realistic growth targets. What are the key factors influencing broker and customer selection of insurers? Who are the key decision-makers at preferred channel partners? What is the best method of communicating and updating risk appetite, value proposition and channel strategy? 4. Defining insurer appetite & capabilities: Potential opportunities are assessed against an insurer’s current and anticipated risk appetite. Priority segments are then compared against an insurer’s capabilities to understand where investment is needed. Inpoint helps its clients manage total exposure while entering new lines. Developing a Customized Growth Program The resulting analysis is used to design a customized growth program outlining the organizational capabilities required to write and service risks in a local market, including: • Defined market focus • Risk appetite communication plan • Detailed distribution program • Budget and capital plan

27


Global Statistics: Motor

5yr LR CV

5yr

3yr

1yr

Cumulative Loss Ratio

Loss Incurred Out (Under) Performance

Annualized Premium Growth

One Year



Three Year



Latest GWP (USD Millions)

Country

Income Per Capita

Five Year

Americas Colombia

Low

1,690

15.8%

599

10.9%

462

15.1%

168

50.9%

52.7%

53.5%

2.6%

Venezuela

Low

5,263

35.9%

1,050

27.2%

673

7.1%

17

60.5%

58.6%

56.8%

5.0%

Puerto Rico

Low

670

(4.6%)

28

(4.7%)

68

(4.1%)

(1)

61.0%

60.1%

61.8%

2.1%

Brazil

Low

14,827

15.2%

318

10.8%

(558)

14.6%

(808)

66.3%

64.8%

62.0%

4.6% 2.0%

U.S.

High

190,044

(0.2%)

82,382

0.3%

53,350

1.5%

15,877

65.9%

63.8%

62.8%

Argentina

Low

3,866

19.9%

(391)

16.0%

(93)

22.9%

(92)

63.2%

64.3%

65.3%

3.6%

Chile

Low

971

16.2%

(156)

14.3%

(108)

26.4%

(42)

65.2%

68.0%

66.9%

2.8%

Mexico

Low

4,494

4.5%

(1,575)

1.9%

(692)

13.4%

(150)

64.2%

69.1%

70.3%

3.7%

Canada

High

17,015

(2.0%)

(3,078)

1.8%

(1,757)

5.9%

80

73.8%

76.9%

75.4%

4.5%

238,842

1.0%

79,177

1.7%

51,346

3.3%

15,049

66.2%

64.7%

63.7%

1.8%

Subtotal Asia Pacific Indonesia

Low

982

15.6%

434

18.5%

315

41.1%

113

49.4%

50.0%

50.2%

2.1%

China

Low

44,361

33.1%

13,862

30.5%

11,221

40.6%

6,695

45.8%

52.2%

52.2%

5.2%

Thailand Hong Kong Taiwan

Low

2,354

14.4%

612

9.9%

484

23.4%

183

53.1%

55.5%

56.1%

1.8%

High

413

4.8%

275

6.7%

237

10.8%

123

44.4%

52.2%

56.1%

8.2% 2.5%

Low

1,664

(1.9%)

326

(0.5%)

201

9.4%

(18)

61.9%

59.2%

58.6%

New Zealand

High

1,058

8.4%

200

8.6%

246

15.7%

134

61.6%

64.3%

67.0%

4.3%

Japan

High

48,767

2.8%

5,776

7.9%

4,135

7.1%

1,318

71.5%

70.2%

68.9%

2.4%

Malaysia

Low

1,844

9.5%

(622)

12.5%

(374)

23.7%

(141)

68.5%

71.1%

71.1%

3.2%

India

Low

4,020

37.4%

(1,532)

10.2%

(1,307)

19.4%

(914)

83.6%

76.1%

72.2%

8.3%

S. Korea

Low

11,768

4.1%

(5,605)

7.5%

(3,853)

14.9%

(1,944)

77.4%

76.0%

73.5%

3.5%

Singapore

High

855

15.9%

(223)

19.8%

(122)

17.0%

(0)

74.3%

78.8%

78.7%

9.5%

Australia

High

11,835

12.0%

(9,377)

14.6%

(4,905)

18.3%

(920)

82.0%

89.7%

92.3%

7.7%

129,923

11.2%

4,126

14.7%

6,279

19.9%

4,628

63.8%

66.6%

66.6%

2.0%

6.9%

Subtotal Europe, Middle East & Africa Ukraine

Low

775

16.8%

498

12.4%

304

2.2%

138

43.0%

46.3%

44.0%

Czech Republic

Low

2,008

5.6%

1,073

1.3%

727

(6.6%)

98

55.9%

52.2%

51.9%

2.7%

Saudi Arabia

Low

864

15.3%

274

9.9%

160

6.0%

24

58.1%

56.5%

54.7%

3.8%

Russia

Low

7,780

13.2%

927

(1.8%)

(660)

8.0%

(611)

68.7%

66.0%

60.2%

11.3%

Hungary

Low

783

(5.6%)

73

(14.9%)

(6)

(16.2%)

(74)

70.3%

63.6%

61.1%

5.1%

Switzerland

High

5,375

4.7%

2,003

5.2%

1,163

5.2%

252

69.5%

65.8%

63.4%

5.3%

Denmark

High

2,776

6.7%

992

1.0%

333

(0.8%)

(18)

74.9%

69.4%

64.1%

9.1%

Austria

High

3,995

2.2%

1,063

(1.3%)

584

7.7%

141

70.7%

68.2%

66.2%

3.1%

U.A.E

High

1,231

15.5%

199

10.5%

185

(6.7%)

85

67.4%

68.3%

67.9%

1.6%

Norway

High

3,258

7.9%

340

5.8%

425

13.7%

308

64.8%

68.4%

69.2%

2.7%

Poland

Low

4,849

8.5%

(1,475)

(1.5%)

(1,128)

14.4%

(186)

64.7%

72.1%

69.2%

6.4%

South Africa

Low

3,224

8.0%

(938)

5.9%

(507)

21.9%

(189)

66.7%

69.4%

69.5%

1.8%

Luxembourg

High

527

7.1%

16

5.5%

4

6.6%

(14)

76.8%

73.0%

70.9%

5.3%

Netherlands

High

5,720

1.0%

101

(1.2%)

(296)

(3.2%)

(71)

75.5%

74.9%

71.2%

5.5%

Spain

High

14,213

0.1%

(864)

(5.6%)

(1,261)

(10.7%)

(397)

77.0%

75.9%

72.7%

4.9%

Finland

High

1,810

3.8%

(120)

0.0%

180

2.5%

193

63.6%

69.9%

72.9%

5.8%

Morocco

Low

841

9.9%

(410)

8.3%

(156)

2.7%

(33)

64.8%

69.8%

73.8%

8.6%

Romania

Low

1,338

4.0%

(1,123)

(14.6%)

(971)

(14.1%)

(378)

89.1%

83.9%

75.7%

11.3%

Portugal

Low

1,999

(4.5%)

(1,455)

(7.8%)

(1,018)

4.4%

(419)

81.8%

80.1%

75.7%

5.7%

Ireland

High

1,896

(2.3%)

(520)

(2.9%)

(723)

(11.0%)

(390)

94.8%

85.0%

76.6%

14.6%

Belgium

High

4,286

3.3%

(1,172)

1.1%

(844)

(1.1%)

19

73.8%

79.7%

77.2%

8.8%

Sweden

High

2,964

(0.1%)

(1,033)

(3.9%)

(153)

3.9%

(97)

77.5%

74.9%

78.2%

5.2%

Low

3,750

8.5%

(2,750)

1.9%

(2,095)

22.3%

(1,044)

88.7%

83.3%

78.8%

8.1%

High

26,433

(0.1%)

(11,244)

(3.5%)

(7,804)

(5.5%)

(2,143)

82.3%

82.5%

79.5%

4.6%

Turkey Italy Iran

Low

3,080

17.2%

(2,042)

14.2%

(1,090)

19.1%

(392)

73.6%

76.7%

79.7%

6.1%

France

High

24,237

1.7%

(14,326)

0.0%

(8,501)

(2.0%)

(2,404)

84.2%

84.6%

83.4%

3.0%

U.K.

High

20,990

(2.7%)

(14,985)

(6.8%)

(10,885)

7.3%

(4,589)

96.1%

90.2%

84.8%

9.3%

Germany

High

26,916

(0.3%)

(33,223)

(1.9%)

(21,516)

(4.2%)

(6,841)

99.7%

98.6%

95.3%

5.1%

Israel

High

2,249

6.3%

(3,182)

1.5%

(2,076)

8.0%

(646)

102.9%

105.7%

102.3%

8.1%

Subtotal

180,167

1.7%

(83,303)

(2.0%)

(57,624)

(0.1%)

(19,677)

82.8%

82.0%

79.2%

4.5%

Grand Total

548,932

3.2%

0

2.8%

0

5.6%

0

71.1%

71.2%

69.8%

2.2%

28

Aon Benfield

Global Statistics: Property

9.6%

144

20.9%

21.1%

21.5%

1.4%

569

20.7%

86

31.5%

27.7%

25.0%

5.6%

Colombia

Low

1,202

14.2%

530

11.2%

482

15.2%

(25)

42.0%

32.4%

34.1%

7.5%

Brazil

Low

8,750

19.5%

472

15.4%

1,248

21.8%

(240)

42.7%

42.1%

43.7%

3.4%

Argentina

Low

1,562

16.4%

(28)

11.0%

84

19.3%

10

39.3%

45.8%

45.7%

4.8%

Mexico

Low

3,166

10.7%

(597)

11.7%

436

26.6%

61

38.0%

42.8%

49.7%

14.8%

U.S.

High

171,915

2.7%

19,824

3.0%

11,874

8.4%

(489)

65.7%

58.3%

57.4%

8.2%

Canada

High

13,576

5.8%

(1,684)

4.4%

(850)

8.9%

(107)

66.2%

62.9%

62.6%

5.9%

Low

1,744

19.8%

(9,247)

14.4%

(9,015)

32.3%

158

30.9%

270.2%

197.4%

329.3%

203,694

3.7%

11,011

20.5%

5,430

9.5%

(401)

63.3%

58.7%

57.6%

6.7% 3.2%

5yr LR CV

602

9.4%

5yr

(1.6%)

844

3yr

897

15.9%

1yr

(1.5%)

1,023


757

Low


Low

Venezuela

Income Per Capita

Puerto Rico

Country




One Year


Three Year



Five Year

Americas

Chile Subtotal Asia Pacific Malaysia Singapore Thailand India

Low

1,168

7.3%

1,522

6.2%

1,116

14.4%

321

12.5%

13.5%

15.4%

High

624

5.5%

757

4.8%

429

9.8%

185

35.7%

36.6%

33.0%

6.1%

Low

841

10.8%

212

13.0%

253

29.6%

59

33.0%

36.0%

38.7%

5.5%

Low

2,023

9.8%

504

7.7%

267

14.1%

(79)

43.9%

42.9%

39.5%

5.2%

Hong Kong

High

712

3.4%

684

3.1%

527

3.7%

263

28.5%

35.2%

39.5%

8.7%

Japan

High

17,116

3.0%

14,986

7.3%

9,521

2.7%

2,487

50.9%

41.4%

40.6%

6.1%

Taiwan

Low

988

(3.7%)

188

(4.9%)

(7)

1.1%

(259)

66.2%

48.2%

41.8%

14.6%

Indonesia

Low

1,429

8.0%

161

7.0%

174

13.6%

(32)

42.2%

43.6%

42.6%

4.4%

S. Korea

Low

2,331

2.9%

(394)

(1.7%)

(266)

13.7%

(218)

49.3%

51.9%

48.7%

7.3%

China

Low

9,789

28.8%

(2,200)

32.8%

(1,117)

21.0%

(569)

45.8%

52.3%

52.0%

5.0%

Australia

High

8,328

10.1%

(2,416)

11.0%

(1,479)

25.2%

541

58.9%

67.5%

67.2%

8.4%

New Zealand

High

1,604

12.0%

(2,340)

13.5%

(2,286)

25.6%

(709)

109.6%

118.9%

98.2%

56.8%

46,953

8.3%

11,665

42.9%

7,133

12.7%

1,989

51.1%

50.2%

48.5%

3.1%


Low

1,261

18.4%

1,770

52.0%

1,510

38.9%

454

4.0%

18.8%

16.3%

11.7%

Russia

Low

5,881

14.5%

6,456

14.5%

4,337

15.1%

1,058

22.0%

19.8%

17.0%

4.9%

Romania

Low

482

10.3%

405

0.8%

238

25.9%

85

22.3%

29.2%

26.2%

6.4%

Iran

Low

731

12.4%

557

3.9%

361

13.1%

93

27.3%

30.1%

27.7%

4.2%

Saudi Arabia

Low

795

13.5%

329

11.1%

237

9.7%

51

33.5%

37.0%

35.2%

9.0%

Turkey

Low

2,235

11.1%

726

1.3%

472

8.0%

(64)

42.9%

40.6%

38.2%

5.5% 11.0%

Hungary

Low

753

4.8%

68

(2.9%)

20

8.2%

(5)

40.7%

47.0%

43.4%

Norway

High

4,278

3.1%

2,833

(1.4%)

1,193

0.6%

479

54.2%

51.1%

46.4%

7.8%

Switzerland

High

4,337

6.6%

2,127

9.7%

1,818

10.8%

1,009

42.1%

45.5%

48.3%

5.3%

Morocco U.A.E

Low

256

8.2%

(37)

5.4%

(8)

3.0%

(45)

57.4%

48.9%

48.4%

9.9%

High

1,227

16.8%

430

12.1%

369

8.3%

137

54.3%

49.6%

51.1%

10.8% 12.5%

Poland

Low

1,731

11.0%

(513)

5.0%

(431)

17.7%

(172)

49.9%

57.5%

52.4%

Portugal

Low

1,043

1.7%

(444)

(1.5%)

(317)

11.1%

(189)

58.1%

58.5%

54.0%

7.3%

Czech Republic

Low

1,021

5.7%

(515)

5.4%

(306)

1.2%

(256)

65.1%

57.8%

56.0%

6.1%

Spain

High

9,320

6.6%

365

2.0%

43

(6.7%)

(44)

65.9%

60.5%

59.1%

4.4%

South Africa

Low

2,783

7.3%

(1,666)

6.3%

(940)

24.4%

(427)

55.3%

61.0%

59.5%

4.7%

Netherlands

High

5,061

4.3%

(114)

1.7%

(281)

(3.8%)

2

65.3%

62.5%

60.4%

4.6%

Belgium

High

2,958

5.8%

(136)

1.9%

(414)

(3.4%)

(147)

70.4%

65.2%

60.8%

8.1%

U.K.

High

24,621

(1.0%)

(1,251)

(4.4%)

1,295

7.8%

773

62.3%

59.0%

60.9%

8.3%

Israel

High

931

6.3%

(180)

4.2%

(127)

9.8%

(15)

67.0%

65.7%

64.3%

3.9%

Italy

High

7,398

1.8%

(2,166)

(1.1%)

(2,154)

(6.9%)

62

64.6%

69.7%

65.5%

7.7%

Sweden

High

3,822

2.6%

(1,198)

(0.6%)

(903)

6.6%

(287)

72.9%

68.5%

66.2%

4.6%

Ireland

High

1,383

(0.8%)

(451)

0.9%

(792)

(5.7%)

(362)

91.5%

79.1%

66.3%

22.6%

Austria

High

3,343

4.2%

(1,288)

(1.7%)

(485)

0.2%

169

60.3%

65.5%

67.6%

9.8%

France

High

25,959

3.6%

(11,258)

1.0%

(7,458)

(2.7%)

(1,435)

70.9%

70.0%

68.7%

4.1%

Finland

High

1,053

0.1%

(699)

(7.2%)

(496)

(16.4%)

(82)

73.2%

74.5%

71.3%

8.2%

Germany

High

22,403

2.3%

(12,719)

1.3%

(6,559)

(3.2%)

(1,622)

72.6%

70.2%

71.4%

5.0%

Luxembourg

High

1,283

32.0%

(756)

12.4%

(408)

(14.6%)

(100)

73.2%

69.7%

71.8%

6.2%

Denmark

High

4,207

7.1%

(3,348)

3.5%

(2,378)

(0.4%)

(708)

82.2%

79.5%

77.0%

8.4%

Subtotal

142,554

3.7%

(22,676)

16.8%

(12,564)

1.2%

(1,588)

63.1%

62.0%

61.4%

3.6%

Grand Total

393,201

4.2%

0

21.0%

0

6.7%

0

61.8%

59.0%

58.1%

5.0%

29


Global Statistics: Liability

5yr

5yr LR CV

33.5%

1,360

59.7%

590

8.3%

8.6%

8.9%

0.8%

1,000

24.5%

720

24.9%

330

24.7%

27.4%

26.9%

2.3%

Mexico

Low

1,434

9.2%

1,879

7.4%

1,131

18.9%

439

33.6%

32.4%

30.3%

4.3%

Brazil

Low

1,819

19.5%

1,879

15.8%

1,420

29.3%

660

27.9%

30.8%

33.5%

5.5%

Low

393

(0.3%)

520

(13.0%)

248

10.9%

51

51.2%

42.1%

38.6%

7.1%

High

5,888

5.6%

2,982

0.8%

1,457

21.1%

324

54.2%

54.2%

50.7%

6.0%

3yr

1,751

22.9%

1yr

33.2%

835


1,055

Low


Low

Colombia

Income Per Capita

Venezuela

Country




One Year


Three Year



Five Year

Americas

Puerto Rico Canada Chile

Low

756

21.4%

196

17.7%

200

33.3%

180

40.5%

51.8%

54.6%

14.6%

Argentina

Low

3,328

25.2%

(52)

22.9%

(76)

38.2%

(48)

65.6%

64.1%

62.6%

3.0%

High

118,694

(2.3%)

(29,160)

(1.8%)

(14,074)

7.7%

(5,790)

64.6%

67.2%

66.5%

7.9%

134,203

(1.1%)

(19,004)

(0.7%)

(7,613)

9.7%

(3,264)

62.5%

65.1%

64.5%

7.2%

U.S. Subtotal Asia Pacific Malaysia

Low

337

8.5%

564

9.8%

390

0.7%

137

23.5%

23.0%

23.5%

1.0%

Indonesia

Low

329

17.8%

396

16.1%

263

26.1%

93

35.9%

32.1%

30.3%

4.9%

High

15,495

0.5%

21,049

7.3%

12,411

6.0%

4,440

31.1%

34.7%

31.5%

5.7%

Low

212

(1.8%)

347

(9.7%)

230

(14.5%)

43

44.0%

33.0%

36.1%

10.1% 11.2%

Japan Thailand India Singapore Taiwan China

Low

1,570

10.3%

1,521

14.7%

802

12.5%

337

42.7%

43.1%

37.4%

High

576

13.6%

492

11.8%

373

12.9%

78

46.2%

40.2%

40.9%

6.5%

Low

307

(5.1%)

273

2.4%

144

13.2%

71

41.0%

46.5%

43.4%

11.2%

Low

3,389

14.7%

1,956

16.2%

1,467

37.9%

997

34.8%

43.8%

45.8%

9.2%

Hong Kong

High

1,023

6.6%

652

6.2%

555

2.6%

236

36.7%

44.8%

47.6%

6.8%

New Zealand

High

295

11.0%

135

1.8%

44

19.3%

(26)

68.5%

57.7%

50.9%

14.1%

Australia

High

7,560

5.8%

3,531

3.5%

1,022

12.8%

571

52.2%

58.2%

51.2%

11.2%

S. Korea

Low

30,893

17.2%

(20,826)

13.0%

(12,883)

37.2%

(5,437)

81.8%

80.5%

80.5%

0.9%

61,986

9.4%

10,089

10.0%

4,817

22.6%

1,538

59.9%

60.1%

57.9%

3.7%

18.9%


Low

620

(15.6%)

2,912

(40.0%)

526

(19.3%)

120

44.9%

40.3%

17.8%

Turkey

Low

326

19.7%

712

2.9%

523

31.0%

149

18.4%

20.9%

21.1%

4.8%

Saudi Arabia

Low

135

3.6%

297

(4.3%)

200

(6.8%)

71

11.4%

15.7%

21.5%

8.0%

Russia

Low

1,049

5.6%

1,750

28.2%

1,127

3.2%

350

30.9%

29.5%

24.3%

10.0%

Poland

Low

1,090

14.3%

1,575

4.5%

942

7.3%

351

32.0%

31.8%

28.9%

4.7%

U.A.E

High

1,097

16.1%

1,414

10.7%

1,007

10.5%

397

23.5%

33.0%

32.1%

8.6%

Norway

3.8%

High

2,878

10.4%

2,918

12.9%

1,980

46.8%

789

32.3%

35.4%

35.4%

Iran

Low

514

33.6%

606

7.3%

467

(15.3%)

137

37.5%

35.5%

36.9%

5.6%

Hungary

Low

193

(1.1%)

247

(10.3%)

129

13.6%

52

37.4%

39.7%

39.0%

2.8%

Czech Republic

Low

781

11.5%

702

5.1%

452

7.5%

137

46.6%

44.0%

42.5%

4.5%

South Africa

Low

785

0.1%

561

3.8%

481

12.0%

166

43.1%

40.5%

46.6%

8.9%

Switzerland

High

2,824

6.2%

1,851

7.9%

1,227

7.9%

453

43.7%

47.7%

46.7%

4.0%

Low

175

(0.4%)

57

8.4%

32

61.1%

52

34.6%

54.8%

54.3%

22.0%

Romania Netherlands

High

3,637

5.4%

1,253

3.5%

837

(1.8%)

234

53.3%

55.7%

54.5%

3.4%

U.K.

High

16,444

(5.3%)

5,713

(9.7%)

2,113

2.2%

1,074

53.2%

59.1%

55.7%

6.3%

France

High

10,728

5.3%

2,251

1.0%

2,180

(3.5%)

313

56.8%

56.6%

57.4%

2.8%

Spain

High

6,839

0.9%

1,575

(7.3%)

260

(13.9%)

465

52.9%

62.0%

57.7%

8.0%

Austria

High

2,255

9.7%

258

6.2%

170

25.1%

17

58.9%

60.2%

58.9%

2.2%

Ireland

High

1,022

(5.5%)

153

(6.3%)

(23)

(15.1%)

(280)

87.1%

63.8%

59.1%

19.8%

Denmark

High

1,038

2.3%

(233)

(5.4%)

51

(12.3%)

(56)

65.1%

61.7%

65.7%

7.1%

Germany

High

15,325

1.9%

(4,409)

(0.5%)

(2,715)

(3.6%)

(1,053)

66.6%

68.8%

67.3%

2.6%

Morocco

Low

379

12.0%

(117)

9.7%

(21)

2.3%

47

51.8%

65.1%

69.3%

15.3%

Portugal

Low

1,079

(3.3%)

(707)

(6.9%)

(277)

5.3%

(74)

71.1%

71.7%

73.9%

3.3%

Italy

High

6,153

1.9%

(4,204)

(3.0%)

(2,635)

(12.1%)

(984)

75.7%

75.9%

74.3%

4.4%

Finland

High

1,234

1.8%

(1,041)

(0.1%)

(566)

14.8%

(117)

69.2%

79.6%

79.5%

6.9%

Luxembourg

High

980

19.6%

(866)

21.1%

(928)

22.0%

(281)

88.4%

96.2%

84.2%

34.0%

Sweden

High

210

2.1%

(204)

15.0%

(182)

47.0%

(128)

120.7%

100.5%

89.1%

36.5%

Israel

High

608

3.7%

(858)

7.7%

(606)

5.8%

(235)

98.4%

97.4%

92.2%

12.3%

Belgium

High

3,049

3.2%

(5,254)

(0.1%)

(3,957)

(4.3%)

(438)

74.1%

104.0%

95.6%

36.1%

83,446

1.4%

8,915

(2.6%)

2,796

(0.9%)

1,726

58.0%

62.1%

59.6%

4.3%

279,634

1.5%

0

0.7%

0

8.8%

0

60.6%

63.1%

61.7%

4.2%

Subtotal Grand Total

30

Aon Benfield

Afterword: The Disappearing Risk For several years we have been using the ratio of premium to GDP as a measure of premium adequacy in the U.S. market. Page 8 explored the ratio more, and explained the insurance cycle as loss driven but generated by pricing and reserving lags. We are now in the eighth year of softening pricing measured by premium to GDP and in 2011 the ratio moved clearly below 3 percent for the first time in over forty years. At the same time the industry reported a combined ratio of 99 percent in the first quarter of 2012, benefiting in part from below average catastrophe losses. Given historically low pricing 99 percent is a surprisingly strong result — albeit representing a return below the industry’s cost of capital in today’s low interest rate environment. How can we reconcile these facts? The graphs to the right show historical frequency indices for several major lines of business. Auto frequency is measured by the rate of fatal accidents per 100 million miles driven. It shows a decline of 60 percent since 1980, or a 3 point average annual decline. Workers’ compensation lost time frequency has declined steadily since 1990 by a total of 55 percent, or a 6 point average annual decline. Medical malpractice frequency has benefited from tort reforms in many states, and shows a decline in the rate of law suits per 100 physicians of 50 percent since 1991, or 4 points per year. General liability shows a similar trend since 1991. Outside the liability sphere, we even see a decline in the number of fires: since 1980 the number of structure fires in the U.S. has declined by more than 50 percent, or more than 2 points per year on average. Frequency is obviously only half the loss cost story; severity is the other half. Severity trends have been broadly stable, with increases within a few points of the CPI or medical CPI rate for most lines over the last decade. Since many lines are rated on an inflation sensitive exposure basis, the resulting premium increases combined with frequency declines have been enough to offset severity increases and have sustained underwriting results through the soft market to a much greater degree than expected. In part this phenomenon is responsible for the favorable reserve development the industry has reported since 2003.

Historical Frequency Trends by Line of Business Auto Fatalities per Miles Driven 2.0 1.5 1.0 0.5 0.0

1980

1985

1990

1995

2000

2005

2010

Workers’ Compensation Lost Time Frequency 2.0 1.5 1.0 0.5 0.0

1990

2000

1995

2010

2005

Medical Professional Liability Frequency 2.0 1.5 1.0 0.5 0.0

1990

1995

2000

2010

2005

Commercial General Liability Frequency 2.0 1.5 1.0 0.5 0.0

1990

1995

2000

2010

2005

Fire Structure Loss Count 2.0 1.5 1.0 0.5 0.0

1980

1985

1990

1995

2000

2005

2010

All charts indexed to 1990 = 1.0

31

Combining all lines of business and comparing industry net incurred loss to GDP produces a very interesting picture (right chart). Loss to GDP was on a steadily rising track from 3.50% 1970 to 1986, increasing from below 2 percent to above 3 percent over that 3.25% period. 1987 marked a sea change in U.S. insurance: a new tax law required discounting incurred losses, making cash3.00% flow underwriting much less attractive; the claims made form and absolute pollution exclusion were 2.75% introduced; and, in casualty lines, we saw the beginning of a trend to limit coverage 2.50% and exclude risk. Since 1987, loss as a percentage of GDP has decreased from 3 percent back to 2 percent. The consistent 2.25% nature of the decline is masked by the clear impact of the reserving cycle as well as random 2.00%in catastrophe losses. The loss to GDP year-to-year variability graph aggregates the impact of favorable frequency trends 1.75% and shows that they have produced a material shift in insurance penetration into the economy. It demonstrates 1.50% the extent to which insurers, all eager to grow, are now fighting over a shrinking cake. The traditionally insured part of the world is becoming safer: loss control programs, health warnings and risk awareness, drunk driving and graduated driving licenses are all reducing the premium and loss size of standard, lower limit liability policies. Against this trend of the “disappearing small risk” we have seen other important trends in recent years. The first has been a trend of increasing property losses driven by catastrophic events. Event frequency and its overlap with high-wealth areas of the globe is illustrated on pages 14–19. Property premium globally has increased by 4.2 percent annually over the last five years, compared to 3.2 percent for motor and just 1.5 percent for liability, the two other major insurance sectors. In the U.S., without premium increases in property of around 15 percent since 2006 the total statutory premium in the industry would have shrunk by much more than the 2 percent drop actually observed. Despite the increase in property exposure, global reinsurance capacity remains more than adequate to meet demand. The peak reinsured exposure remains U.S. hurricane, followed by U.S. earthquake, and then European and Japanese wind and earthquake exposures (pages 18–19). The amount of risk transferred into the private reinsurance market is greatly reduced by a number of (implicit or explicit) government pools and public sector solutions, most notably the non-insurance of earthquake in the U.S. and the Japan Earthquake Reinsurance pool. Many of these pools were created when reinsurance capacity was lower and not

32

Calendar Year Incurred Loss % of GDP 3.50% 3.25% 3.00% 2.75% 2.50% 2.25% 2.00% 1.75% 1.50% 1970 1975 1980 1985 1990 1995 2000 2005 2010 1965

1970

1975 1980 1985 1990 1995 2000 2005 2010

deemed sufficient to meet private market demand. But with 1965 1970 levels 1975 of 1980 1985 1990 1995 2000 2005 2010to today’s record capacity, the opportunity exists move substantial risk into the private insurance and reinsurance markets. Ballooning public sector debt is further reason why reliance on pre-funded reinsurance solutions is clearly a superior risk management solution. Another important trend has been the emergence of corporate liability losses on a scale that would have been hard to imagine just five years ago. Companies in the oil drilling, power generation and pharmaceutical industries, as well as numerous examples in financial services, have suffered actual or market capitalization losses of tens of billions of dollars or more over the last two years. Almost all of these losses have been uninsured. Many other potential exposures of a similar size exist in the market: railway liability from an explosion or derailment of dangerous cargo in an urban area, the multi-channel accumulation of BPA through the bio-system, damage from solar storms, and breach of privacy related to social media or other cyber-crime to name a few. In fact, the insurance world is mirroring the financial and manufacturing worlds. Systems are becoming more reliable on average, and less susceptible to minor disruptions, but have an increased fragility and exposure to major catastrophic losses. Just-in-time inventory systems and globally distributed manufacturing led to some surprising business interruption claims from the Tōhoku earthquake, for example. And the contagions in the financial crisis of 2007-09 are well known.

2002

2000

2001

1998


1999

1997

1996

1995

1994

1992

1993

1991

1991

1990

1989

1988

1987

1986

1985

1984

1983

1982

1981

0.0

1980

0.2

Aon Benfield

Just as insurance capacity has been deployed to effectively transfer and manage property risks once thought too large, the industry today needs a similar expansion of capacity for casualty and non-traditional exposures to help drive meaningful growth over the next decade. In moving to self-insurance or no insurance, an important independent oversight function is lost: the loss control and process review of the risk taker. Such a service, backed by a risk indemnity, has been so effective for small risks that they are rapidly decreasing. To reverse the resulting long-term decline in the insurance growth and decreasing loss to GDP, the industry needs to creatively and constructively grapple with large, harder to quantify risk. What lessons can we draw from the successful transfer of mega-property risks to insurers and reinsurers for potential casualty capacity? Analytics and catastrophe models apply science and statistics to the problem of estimating losses. Models provide a common currency for risk — a currency shared by risk assumers, regulators, rating agencies, and, crucially, capital providers — and they support market pricing that is transparent and predictable. The same computer modeling techniques used so successfully to manage property risk are now developing to the point where some mega-casualty perils could be understood well enough to support insurance solutions — provided there is adequate demand.

The global pooling provided by reinsurance (whether traditional or non-traditional capital markets) is key to creating capacity: reinsurance provides property capacity for a number of independent but highly volatile risk towers that together drive adequate, but manageable, leverage of capital and so adequate returns on capital. Success in providing higher capacity to casualty and other nonproperty risks will not be based on one or two highly skewed covers, but on an adequate portfolio of such covers. Success will feed on itself, with more risk attracting more capital and driving more cost effective coverage. The first steps will come from an open attitude to risk from insurers, and, equally importantly, a greater appreciation for the value of the insurance product from a range of insureds. Aon Benfield continues to work with its clients and reinsurance markets to help facilitate a creative and responsible approach to risk innovation through new products, often with reinsurance support, to drive insurance growth in the coming decade.

33


Sources: Global Risk Parameters ANIA (Italy), Association of Vietnam Insurers, BaFin (Germany), Banco Central del Uruguay, Bank Negara Malaysia, CADOAR (Dominican Republic), Cámara de Aseguradores de Venezuela, Comisión Nacional de Bancos y Seguros de Honduras, Comisión Nacional de Seguros y Fianzas (Mexico), Danish FSA, Dirección General de Seguros (Spain), DNB (Denmark), E&Y Annual Statements (Israel), Finma (Switzerland), FMA (Austria), FSA (U.K.), HKOCI (Hong Kong), http://www.bapepam.go.id/perasuransian/index.htm (Indonesia), ICA (Australia), Insurance Commission (Philippines), IRDA Handbook on Indian Insurance Statistics, Korea Financial Supervisory Service, Monetary Authority of Singapore, MSA Research Inc. (Canada), Quest Data Report (South Africa), Romanian Insurance Association, Slovak Insurance Association, SNL Financial (U.S.), Superintendencia de Banca y Seguros (Peru), Superintendencia de Bancos y Otras Instituciones Financieras de Nicaragua, Superintendencia de Bancos y Seguros (Ecuador), Superintendencia de Pensiones de El Salvador, Superintendencia de Pensiones, Valores y Seguros (Bolivia), Superintendencia de Seguros de la Nación (Argentina), Superintendencia de Seguros Privados (Brazil), Superintendencia de Seguros y Reaseguros de Panama, Superintendencia de Valores y Seguros de Chile, Superintendencia Financiera de Colombia, Taiwan Insurance Institution, The Insurance Association of Pakistan, The Statistics of Japanese Non-Life Insurance Business, Turkish Insurance and Reinsurance Companies Association, Yearbooks Of China’s Insurance, and annual financial statements U.S. Risk Parameters SNL Financial U.S. Reserve Adequacy and Risk SNL Financial, A.M. Best Modeling the Underwriting Cycle A.M. Best, U.S. Bureau of Economic Analysis Motor Insurance in China China Statistical Yearbooks, Yearbooks of China’s Insurance A Different View of Risk Axco Insurance Information Services, EM-DAT International Disaster Database, Global Seismic Hazard Assessment Program (GSHAP), Impact Forecasting (Aon Benfield), International Best Track Archive for Climate Stewardship (IBTrACS), LandScan 2006™ High Resolution global Population Data Set, Penn World Table Version 7.0, USGS EROS GTOPO30 global digital elevation model, World Bank 2005 The Wealth of Nations Dataset Global Crop Insurance Penetration USDA Foreign Agricultural Service Correlation Considerations FSA (U.K.), SNL Financial (U.S.), The Statistics of Japanese Non-Life Insurance Business, Yearbooks Of China’s Insurance Macroeconomic Correlation Bloomberg, Case-Schiller U.S. National Home Price Index, U.S. Bureau of Economic Analysis, U.S Bureau of Labor & Statistics, Yahoo! Finance Global Market Review Axco Insurance Information Services, CIA World Factbook July 2012 Population Estimates, IMF World Economic Outlook Database April 2012 Edition Afterword A.M. Best, Insurance Services Office (ISO), National Council on Compensation Insurance, Inc. (NCCI), National Fire Protection Association (NFPA), National Highway Traffic Safety Association (NHTSA), National Practitioner DataBank, U.S. Bureau of Economic Analysis

© 2012 Aon Benfield. This document is intended for general information purposes only and should not be construed as advice or opinions on any specific facts or circumstances. The comments in this summary are based upon Aon Benfield’s preliminary analysis of publicly available information. The content of this document is made available on an “as is” basis, without warranty of any kind. Aon Benfield disclaims any legal liability to any person or organization for loss or damage caused by or resulting from any reliance placed on that content. Aon Benfield reserves all rights to the content of this document.

Scan here to access prior editions of the Insurance Risk Study.

34

Aon Benfield

About the Study Rating agencies, regulators and investors today are demanding that insurers provide detailed assessments of their risk tolerance and quantify the adequacy of their economic capital. To complete such assessments requires a credible baseline for underwriting volatility. The Insurance Risk Study provides our clients with an objective and data-driven set of underwriting volatility benchmarks by line of business and country as well as correlations by line and country. These benchmarks are a valuable resource to CROs, actuaries and other economic capital modeling professionals who seek reliable parameters for their models. Modern portfolio theory for assets teaches that increasing the number of stocks in a portfolio will diversify and reduce the portfolio’s risk, but will not eliminate risk completely; the systemic market risk remains. This behavior is illustrated in the left chart below. In the same way, insurers can reduce underwriting volatility by increasing account volume, but they cannot reduce their volatility to zero. A certain level of systemic insurance risk will always remain, due to factors such as the underwriting cycle, macroeconomic trends, legal changes, and weather. The Study calculates this systemic risk by line of business and country. The Naïve Model on the right chart shows the relationship between risk and volume using a Poisson assumption for claim count — a textbook actuarial approach. The Study clearly shows that this assumption does not fit with empirical data for any line of business in any country. It will underestimate underwriting risk if used in an ERM model.

Asset Portfolio Risk

Insurance Portfolio Risk

Portfolio Risk

Insurance Risk

Portfolio Risk

Systemic Insurance Risk

Systemic Market Risk

Naïve Model

Number of Stocks

Volume

For more information on the Insurance Risk Study or our analytic capabilities, please contact your local Aon Benfield broker or:

Americas

Asia Pacific

Brian Alvers +1 312 381 5355 [email protected]

Rade Musulin +61 2 9650 0428 [email protected]

Stephen Mildenhall Chief Executive Officer Aon Benfield Analytics +1 312 381 5880 [email protected]

Europe, Middle East and Africa

John Moore Head of Analytics, International +44 (0) 20 7522 3973 [email protected]

U.K.

Carole Ho +852 2862 4183 [email protected]

Paul Kaye +44 (0) 20 7522 3810 [email protected]

Ben Miliauskas +61 2 9650 0431 [email protected]

Marc Beckers + 44 (0) 20 7086 0394 [email protected]

George Attard +65 6239 8739 [email protected]

35

200 E Randolph Street Chicago Illinois 60601 t: +1 312 381 5300 | f: +1 312 381 0160 | aonbenfield.com Copyright Aon Benfield Inc 2012 | #10032 - 08/2012