Natural Disaster Risk and International Financial Markets:
An Investigation of the Relationship between the Risk of Extreme Weather Events and its Potential Impact on Risk Premia on Country- and Industry-Level
Master Thesis in International Financial Management at
the Faculty of Economics and Business at
the University of Groningen, the Netherlands
Supervisor: Dr Ambrogio Dalò
Word count: 19,905
Student number: S3517330 Name: Anna Andrae
Study Programme: MSc International Financial Management
Field Key Words: market risk premium, natural disaster risk, pricing of risk, rare events
Table of Contents
I. List of Abbreviations ... D II. List of Symbols ... D III. List of Tables ... E IV. List of Figures ... F
1. Introduction ... 1
2. Literature and Hypotheses ... 3
2.1. Fundamental Concepts in Finance Theory ... 3
2.2. Existing Literature on Rare Disasters ... 5
2.3. Derivation of Hypotheses ... 6
3. Data ... 9
3.1. Natural Disaster Risk in light of the World Risk Index ... 9
3.2. Data regarding the Different Markets ... 14
4. Methodology ... 16
4.1. Baseline Model ... 16
4.1.1. Derivation of the Baseline Model ... 16
4.1.2. Limitations of the Baseline Model ... 18
4.2. Model regarding Coping Abilities ... 19
4.3. Model regarding Industry Specifications ... 21
5. Empirical Results and their Discussion ... 24
5.1. Descriptive Statistics and Correlation Matrices providing first Insights 24 5.1.1. Descriptive Statistics ... 24
5.1.2. Correlation Matrices ... 30
5.2. Empirical Results following the Regression Analyses ... 31
5.2.1. Japan ... 31
5.2.2. The Netherlands ... 37
5.2.3. The United States ... 42
5.2.4. Country Comparison of Results to obtain an Integrated View ... 48
5.3. Robustness Checks ... 50
6. Conclusion: Drawing Managerial Implications from the Results ... 54 V. Appendix ... G
A-1 Robustness Check using Fama-French Risk Factors and Momentum Factor ... G A-2 Robustness Check using Fama-French Risk Factors and Momentum Factor in Combination with Macro-economic Control Variables ... J VI. References ... M
I. List of Abbreviations Abbreviation Meaning
CAPM Capital Asset Pricing Model cf. conferatur; compare
e.g. exempli gratia; for example et al. et alia; and others
FTSE Financial Times Stock Exchange GDP Gross Domestic Product
i.e. id est; that means
IPCC Intergovernmental Panel on Climate Change ISIN International Securities Identification Number NAICS North American Industry Classification System
OECD Organization of Economic Cooperation and Development p page
pp pages
UNFCCC United Nations Framework Convention on Climate Change US United States
II. List of Symbols Symbol Meaning
𝛼 significance level
𝑒𝑥𝑝𝑜𝑠𝑢𝑟𝑒 exposure-score as measured in the World Risk Report 𝐹𝑋 foreign exchange rate to US Dollar
𝐺𝐷𝑃𝑔 growth in Gross Domestic Product HML high minus low factor
𝐼𝑁𝐹𝐿 inflation rate
𝐼𝑁𝑇𝑅 short-term interest rate 𝑙𝑎𝑐𝑘 𝑜𝑓 𝑎𝑏𝑖𝑙𝑡𝑖𝑒𝑠
𝑡𝑜 𝑎𝑑𝑎𝑝𝑡 lack of adapting abilities-score as measured in the World Risk Report 𝑙𝑎𝑐𝑘 𝑜𝑓 𝑎𝑏𝑖𝑙𝑖𝑡𝑖𝑒𝑠
𝑡𝑜 𝑐𝑜𝑝𝑒 lack of coping abilities-score as measured in the World Risk Report 𝜆1
̂ estimated market risk premium 𝑀𝐺3 broad money growth measure M3
𝑁 number of observations or given number 𝑅𝑖𝑒 excess return of individual asset
𝑅𝑚𝑒 excess return of market
𝑅̅𝑖 average return of individual asset Rf risk-free rate
SMB small minus big factor
𝑠𝑢𝑠𝑐𝑒𝑝𝑡𝑖𝑏𝑖𝑙𝑖𝑡𝑦 susceptibility-score as measured in the World Risk report 𝑇 measure of time; number of days, months or years
𝑢𝑖, 𝑣𝑖, 𝜀𝑖 residuals
𝑣𝑢𝑙𝑛𝑒𝑟𝑎𝑏𝑖𝑙𝑖𝑡𝑦 vulnerability-score as measured in the World Risk report
III. List of Tables
Table 1: Definition and sub-categories of the components of the World Risk Index. ... 12
Table 2: Descriptive statistics of the variables used in the analyses for Japan. ... 26
Table 3: Correlation matrix of the variables used in the analyses for Japan. ... 26
Table 4: Descriptive statistics of the variables used in the analyses for the Netherlands. ... 27
Table 5: Correlation matrix of the variables used in the analyses for the Netherlands. ... 27
Table 6: Descriptive statistics of the variables used in the analyses for the United States. ... 28
Table 7: Correlation matrix of the variables used in the analyses for the United States... 28
Table 8: Regression output of analyses for Japan as outlined in Section 4. ... 33
Table 9: Regression output of analyses for Japan as outlined in Section 4.3. ... 35
Table 10: Regression output of analyses for the Netherlands as outlined in Section 4. ... 38
Table 11: Regression output of analyses for the Netherlands as outlined in Section 4.3. ... 40
Table 12: Regression output of analyses for the United States as outlined in Section 4. ... 44
Table 13: Regression output of analyses for the United States as outlined in Section 4.3. ... 46
Table 14: Summary of results regarding the stated hypotheses. ... 50
Table 15: Regression output of analyses for Japan as outlined in Section 5.3 using the HML measure to estimate betas. ... 51
Table 16: Regression output of analyses for the Netherlands as outlined in Section 5.3 using the HML measure to estimate betas. ... 52
Table 17: Regression output of analyses for the United States as outlined in Section 5.3 using the HML measure to estimate betas. ... 53 Table 18: Regression output of analyses for Japan as outlined in Section 5.3 using the Fama- French risk factors as controls. ... G Table 19: Regression output of analyses for the Netherlands as outlined in Section 5.3 using the Fama-French risk factors as controls. ... H
Table 20: Regression output of analyses for the United States as outlined in Section 5.3 using the Fama-French risk factors as controls. ... I Table 21: Regression output of analyses for Japan as outlined in Section 5.3 using the Fama- French risk factors in combination with the macro-economic variables as controls. ... J Table 22: Regression output of analyses for the Netherlands as outlined in Section 5.3 using the Fama-French risk factors in combination with the macro-economic variables as controls. ... K Table 23: Regression output of analyses for the United States as outlined in Section 5.3 using the Fama-French risk factors in combination with the macro-economic variables as controls. L
IV. List of Figures
Figure 1: Composition of World Risk Index and its components. ... 13
1 1. Introduction
The relationship between risk and return is one of the fundamental concepts in the domain of Finance, because it provides for example the basis for capital market models and investment decisions. However, translating the simple concept into practice yields substantial difficulties since the variety of risks is exceptional. There are financial risks, like capital risk, market risk or liquidity risk, as well as non-financial risks, like IT risk, compliance risk or conduct risk (Walter et al., 2017). A rarely stated aspect is natural disaster risk, despite its increasing impact on the world’s population due to possible losses, poverty and fatalities (World Bank, 2019).
Against the background of climate change, extreme natural events leading to disasters, such as tornados, tsunamis, wildfires or flooding, may become more frequent and/or more extreme when they occur (IPCC, 2012). Therefore, the anticipation of natural disaster risk also in a financial context and the inclusion of this risk in the pricing of assets gains importance (cf.
OECD, 2015; Wolfrom and Yokoi-Arai, 2016).
In recent years, numerous extreme weather events and disasters were experienced around the world. Examples are Japan’s most powerful earthquake ever recorded in 2011 (BBC, 2011), California’s deadliest wildfire in 2018 (The New York Times, 2018), and the Sumatra- Andaman earthquake and subsequent tsunami devastating parts of Indonesia, Sri Lanka, Thailand and India (National Science Foundation, 2005). In all these disasters lives were lost and extensive losses were recorded due to the destruction of infrastructure, housing and property. Often, humanitarian aid in form of hunger aid (BBC, 2005; World Food Programme, 2005) and search, medical and rescue teams (see German Aerospace Center, 2011; Hindustan Times, 2011; Nebehay, 2011), as well as philanthropic donations (see Japan Center for International Exchange, 2016) are provided in response to such disasters. Besides these actions to help the regions and countries to initially respond, policy makers and governmental players often provide monetary help and policy reliefs whenever possible and effective (Penn and
2 Eavis, 2019). The economic perspective on these events is mostly limited to the calculation of the total economic damage and loss created by the disaster. Worthington (2008) is strengthening this notion by finding that natural events and disasters do not have a significant impact on stock returns at a market level in Australia. Having observed massive devastation as consequence of natural disasters multiple times, the question arises if investors are considering the risk of these natural disasters when making investments decisions. This is what this thesis tries to examine by answering the research question
Is natural disaster risk priced in international stock markets?
To examine the impact of natural disaster risk on market risk premia, first, a Fama- MacBeth regression (see further Fama and MacBeth, 1973) is performed. Following this, it is examined if a country’s ability to cope with disaster influences the before-named relationship.
Lastly, the possible influence of natural disaster risk on the risk premium is examined on an industry level to test if different industries are diversely affected.
Overall, weak support that natural disaster risk is considered in international financial markets is found. Additionally, the vulnerability of a country does influence the market risk premium, but it does not always increase the demanded risk compensation by investors. Further, the results suggest an impact of natural disaster risk on the industry-specific risk premium, although it strongly varies across industries and countries. These findings lead to the conclusion that the anticipation as well as the judgement of natural disaster risk and the vulnerability of a country varies across financial markets around the world leading to a varying impact on risk premia.
This research contributes to the existing literature in three ways. The examination of pricing of natural disaster risk expands the understanding of the functioning of financial markets and gives insights into which risks are considered by investors. Further, the differentiation of the effect on different industries leads to managerial implications for executives active in these industries regarding (re-)location and diversification decisions. Third, the present thesis
3 addresses a gap in the existing literature in terms of shedding light on the relatively young area regarding the financial effect of natural disaster risk and the increasing importance of the topic as outlined above.
The subsequent research is structured as follows: Section 2 reviews the relevant literature in the field and derives testable hypotheses. Subsequently, the data and methodology are examined in Sections 3 and 4, before the empirical results are presented and discussed in Section 5. This study ends with concluding remarks and practical implications in Section 6.
2. Literature and Hypotheses
2.1. Fundamental Concepts in Finance Theory
In the area of finance, risk is a central subject. Finding an overarching definition of the term risk has proven to be difficult (see March and Shapira, 1987; Yates and Stone, 1992), however, risk is understood as “to expose [someone] to the chance of injury or loss” (Little et al., 1983, p. 1743) or other adverse or unwelcome circumstances. Yates and Stone (1992) suggest that there are three components of risk: the potential losses, the significance of these losses, and the uncertainty of these losses, which indicate that risk is closely related to uncertainty.
Based on human perception of uncertainty, one of the fundamentals of finance theory is the relationship between risk and return as laid out by Merton (1980). Under the assumption that, in general, investors are risk-averse and rational, the demanded compensation for an investment is positively related to the risk of the same. Thus, riskier investments are expected to yield higher returns to compensate risk averse investors (Merton, 1980). However, when considering the relationship between risk and return, the distinction between systematic and unsystematic risk is essential (see Beja, 1972; Sharpe, 1964). Systematic risk is “perfectly correlated among all securities“ (Beja, 1972, p. 37) and, therefore, affects the entire market.
Unsystematic risk, in contrast, is not correlated neither between securities, nor with the
4 systematic risk. Hence, it can be interpreted as being asset-specific (Beja, 1972). Following the proposition outlined by Beja (1972, Equation 3, p. 38), it becomes evident that the rate of return of an individual security can be divided into two components: (A) a systematic component and (B) an unsystematic component or residual. According to the author, the risk premium (i.e., the additional compensation for taking on higher risk) is proportional to component (A), but fully independent of component (B) (Beja, 1972). Hence, the unsystematic risk component (B) does not affect the expected return of an investment, leading to no additional compensation. In turn, this yields to the conclusion that investors only get rewarded for taking on systematic but not unsystematic risk.
The concept of diversification addresses exactly this insight. Markowitz (1952) laid out that investors can and should reduce the variance of the expected returns of their investments by investing in multiple assets with a low covariance among them. Further, Sharpe (1964, p.
441) concludes that “diversification enables the investor to escape all but the risk resulting from swings in economic activity – this type of risk remains even in efficient combinations”, indicating that market-specific, systematic risk, in contrast to all other and including unsystematic risk, cannot be eliminated through diversification.
In line with the notion that risk is closely connected with uncertainty, other classifications of risk on basis of various origins of uncertainty are discussed. In an economic context, it is commonly distinguished between the two main types of financial and non-financial risk of which each incorporates various sub-classifications. Examples for financial risks are capital risk, market risk or liquidity risk, whereas operational risk, compliance risk, cyber risk, IT risk, conduct risk or third-party risk are classified as non-financial risks (Walter et al., 2017). As the economic world is rather complex and interwoven (World Economic Forum, 2019), some risks may be relevant for individual parties in the economic systems but not for others, whereas some risks might be the same for all. Hence, it appears to be rather difficult, even impossible, to
5 assemble an all-encompassing cluster of risks and to quantify all of them (cf. World Economic Forum, 2019).
However, in financial markets, the market risk premium quantifies the risk premium for the respective market. In detail, the market risk premium refers to the expected return of the market portfolio, meaning a portfolio comprised of all financial instruments in the market including equity minus the rate of return for an investment without risk, which is the risk-free rate (see further, Sharpe, 1964; Lintner, 1965). Consequently, it gives an indication of how much compensation investors require for investing in the respective market. Drawing on the concept of systematic and unsystematic risk as outlined above, the market risk premium should be based on systematic risks only, as unsystematic risks are diversifiable. Therefore, the market risk premium should incorporate all systematic risks present in the market regardless of their origin or form, even if it cannot be identified which exact risks are at play and what their contribution to the risk premium in the market is.
2.2. Existing Literature on Rare Disasters
A body of literature deals with rather improbable but high-impact risks, such as rare disasters.
Rietz (1988) lays out the basis for this field as he finds that possible but unlikely economic crashes might be one explanation for higher equity risk premia, suggesting that these events do influence the risk premium of an asset. Barro (2006) extends this model by looking at international economic disasters and introducing states of probabilities for entering a disaster, in particular, wars or depressions. He finds that this probability indeed explains a proportionate change in the Gross Domestic Product (GDP), regardless of the length of the disaster. Instead of a probability of disaster that remains constant over time, Gabaix (2012) introduces a time- varying model which examines the effect of a time-varying (economic) disaster severity on stock and bond prices. Looking beyond probabilities and into the real effects of disasters, Nakamura et al. (2013) examine disasters and allow for a recovery period. By studying
6 consumption disasters, such as World Wars, economic collapses, and financial crises, the authors show that after five years about half of the consumption drop caused by a disaster is reverted. They, too, find that disaster risk influences the equity premium, although the magnitude of the influence is reported smaller than in previous studies (Rietz, 1988; Barro, 2006). Time-varying probabilities of consumption disasters, which technically can be seen as a left tail risk, are modelled by Wachter (2013). Her results suggest that higher left tail risk leads to higher equity premia and that the magnitude of this influence is higher than in static models.
Furthermore, Barro and Jin (2016) examine the relationship between rare events and long-run risks on equity risk premia. They find that changes in the assessed probability of rare events can explain alterations in these risk premia, implying that only the probability of a disaster, even without an actual occurrence of such an event, impacts the equity risk premia.
When taking investors’ fear into account, it is found that the US stock market “generally incorporates possible occurrence of rare disasters in the way it prices risky payoffs” (Bollerslev and Todorov, 2011, p. 2166). Further research also highlights that jumps in stock prices can determine a conditional equity premium (Maheu et al., 2013; Guo et al., 2014). Except for Berkman et al. (2017), most of these studies characterise rare disasters as a small probability for a big drop in consumption or the GDP, mostly around 10%. In contrast, Berkman et al.
(2017, p. 351) use a political instability measure as a proxy for rare disaster risk and find evidence in line with previous research (see Bollerslev and Todorov, 2011; Wachter, 2013;
Barro and Jin, 2016), because a “time-varying probability of rare disasters helps to explain fluctuations in expectations of the equity risk premium“.
2.3. Derivation of Hypotheses
The fundamental concepts as well as existing literature in the field of finance suggest that the consideration of risks is important, especially, when interacting in financial markets. That is because risks affect the expected and required returns and, therefore, investment decisions.
7 In sum, results of the existing body of literature (cf. Rietz, 1988; Bollerslev and Todorov, 2011; Wachter, 2013; Berkman et al., 2017) indicate that the probability of economic disasters has a proportional impact on equity risk premia. However, defining rare events only in the context of economic measures may not be suitable for all risks. When broadening the perspective beyond direct economic risks, several other sources and types of risks, for instance the probability of “loss of biodiversity, life, health, culture, identity or social cohesion” (Bread for the World – Protestant Development Service, 2017, p. 15; see further, World Economic Forum, 2019) appear. A key origin of these potential non-economic losses is natural disaster risk (UNFCCC Secretariat, 2014).
Natural disaster risk does not seem to be considered yet in a financial context, despite its increasing impact on the world’s population due to possible losses, poverty, and fatalities (World Bank, 2019). Against the background of climate change, extreme natural events leading to disasters, such as tornados, tsunamis, wildfires or flooding, might become more frequent and/or more intense (IPCC, 2012). Therefore, the anticipation of natural disaster risk, also in a financial context and the inclusion of this risk in the pricing of assets, gains importance (cf.
OECD, 2015; Wolfrom and Yokoi-Arai, 2016).
To date, there is a gap in the literature examining if the risk of extreme natural events leading to natural disasters is considered in the pricing mechanisms in financial markets.
Furthermore, it is not clear yet, if natural disaster risk can be classified as systematic or unsystematic risk. Considering that extreme natural events are independent of the single economy or market, and therefore, should rather affect a given market as a whole, it seems plausible that natural disaster risk could be classified as a systematic risk. If that is the case, one would expect natural disaster risk to be included in the market risk premium among other systematic risks. However, this relationship cannot be subsumed, because the types of risks included in the market risk premium are not directly observable. Conversely, contemplating the impact of natural disasters, it seems reasonable to assume, that different companies or industries
8 are affected by natural disasters in various ways due to different operations and core business activities. Hence, it seems plausible that natural disaster risk could be classified as unsystematic risk, too. Following this line of thought, due to possible natural disasters, single industries or companies would be facing higher risks than others. Consequently, they should have a higher industry risk premium or company-specific risk.
Both options of natural disaster risk being a systematic or an unsystematic risk have not been examined by scholars yet. To address this gap in the literature, this thesis draws on the findings of the existing literature which indicates that disaster risk, proxied by the probability of economic disasters, positively influences the equity risk premium (e.g., Bollerslev and Todorov, 2011; Nakamura et al., 2013; Wachter, 2013). Based on this observation, it can be suggested that the probability of an extreme natural event leading to a natural disaster has a similar influence on the market risk premium. This assumption seems valid as the equity risk premium is closely related to the market risk premium (cf. Donaldson et al., 2010). Hence, Hypothesis 1 is stated as:
H1: The risk of extreme natural events increases the market risk premium.
It is crucial to note that the occurrence of an extreme natural event does not necessarily lead to, but may result in, a natural disaster. The magnitude of the natural disaster might be highly dependent on the vulnerability of a country regarding extreme natural events (Bündnis Entwicklung Hilft, 2011). Despite risk exposure, susceptibility as well as abilities, or the lack thereof, to cope with a disaster and to adapt to changing demands are components of the overall disaster risk (IPCC, 2012; Bündnis Entwicklung Hilft, 2018). The latter three components can be grouped under the generic term ‘vulnerability’ (cf. Bündnis Entwicklung Hilft, 2018, p. 36).
Hence, it could be insightful to observe whether vulnerability has a significant influence on the pricing of natural disaster risk and, if so, which component is a significant driver. Since better
9 coping abilities and adaption to new circumstances are expected to increase flexibility and alleviate the long-term impact of natural events, Hypothesis 2 is stated as:
H2: A country’s vulnerability regarding extreme natural events increases the market risk premium.
As the before-outlined evidence only examines the influence of disaster risk at an economy level (i.e., macro level), the question has not yet been addressed on the level of industries or companies, (i.e., micro level). Practitioners (e.g., Ralph, 2019; Ralph and Keohane, 2019;
Jenkins, 2019) and academic scholars (cf. Loayza et al., 2012; Fomby et al., 2013; Ulubaşoğlu et al., 2019) alike provide evidence that different industries are affected by extreme natural events to a different degree. Industries that are repeatedly stated to be influenced by natural events and subsequent disasters are the Manufacturing, Agriculture (Guidry and Pruitt, 2012;
Loayza et al., 2012; Ulubaşoğlu et al., 2019), and the Insurance industry (Jenkins, 2019).
However, it is noteworthy that these industries have different sensitivities towards their vulnerability depending on the type of natural events and disasters, and the timing of the same (Jenkins, 2019; Ulubaşoğlu et al., 2019). Based on these differences, Hypothesis 3 is stated as:
H3: The effect of natural disaster risk on the industry-sector risk premium varies across industries.
3. Data
3.1. Natural Disaster Risk in light of the World Risk Index
To be able to test the proposed hypotheses, the scope of the analyses comprises three international financial markets, namely the United States, the Netherlands, and Japan.
Although, these countries are located on different continents and have different magnitudes of natural disaster risk (Bündnis Entwicklung Hilft, 2011), all countries studied are advanced
10 economies (International Monetary Fund, 2016). Thus, the three countries are comparable which increases the external validity of the study.
This study is centred around natural disaster risk. Therefore, the independent variable is the World Risk Index score, which subsequently is decomposed in its components. Hence, the exposure to natural disasters and the components of vulnerability are used as independent variables. To retrieve a dataset for the analysed countries, the World Risk Report1 is used as a data basis. Since 2011, a World Risk Report has been published every year by Bündnis Entwicklung Hilft in cooperation with the Ruhr University Bochum – Institute for International Law of Peace and Armed Conflict and the United Nations University, Institute for Environment and Human Security situated in Bonn, Germany. For the analyses, the reports of the years 2011 to 2016 are used, which all have a similar structure. Within the reports, an overall natural disaster risk measure as well as measures for the single components, such as (1) susceptibility, abilities, or the lack thereof, (2) to cope with a disaster and (3) to adapt to changing demands, are calculated for 173 countries. These measures are, then, organised in a ranking. Further, the measures for the three components are aggregated in a score termed ‘vulnerability’. More detailed information on the definition of the measures and their construction are compiled in Table 1 and Figure 1 (see pp. 12). In Figure 1, all components with their respective sub- categories are shown. For each sub-category the variables used for the calculation are listed.
The sub-categories Housing Conditions, Disaster preparedness and early warning, Social networks, and Adaption strategies are blacked out as for these, no (relevant) data could be gathered according to the publishers of the report. The fraction in front of each item represents the weight assigned to the item when calculating the measure of the level above (e.g., the measure for Education and research is calculated by 1
2 Adult literacy rate + 1
2 Combined gross school enrolment).
1Via: www.worldriskreport.org
11 The authors of the reports use different high-quality data sources, for instance, the Global Risk Data Platform PREVIEW of the United Nations Environment Programme, data by the Center for Remote Sensing of Ice Sheets of the University of Kansas, and data by the Center for International Earth Science Information Network of the Columbia University (see Bündnis Entwicklung Hilft, 2011), to construct their measures. As this data is grounded in academia and/or composed by well-known, independent supra-national organizations, the data can generally be regarded as high-quality and up to date (cf. Wurgler, 2000; Fratzscher and Imbs, 2009; Nakamura et al., 2013).
The World Risk Reports for the periods 2011 to 2016 include yearly data from which the overall natural disaster risk score data and scores on its single components, as well as the aggregated vulnerability measure for the US, the Netherlands, and Japan can be retrieved. This data is extracted and compiled in a table categorising information for Country, World Risk Index, Exposure, Vulnerability, Susceptibility, Lack of coping capacities, and Lack of adaptive capacities in separate columns. As monthly data is needed, the yearly values are copied to all calendar months of the respective year. The measures are formatted as percentage numbers, meaning that a value of 5.89% is stated as 5.89 to match the formats of the risk-free rate as well as the market and stock returns, and are converted to a txt-file.
Component Definition Sub-categories Aggregation level Exposure “[…] refers to entities exposed and prone to
be affected by a hazard event.” (p. 15)
− Earthquakes
− Storms
− Floods
− Droughts
− Sea level rise
Exposure
Susceptibility “ […] the likelihood of suffering harm and damages in case of the occurrence of a natural hazard.“ (p. 16)
In the context of the study, “susceptibility refers to selected structural characteristics of a society and the framework conditions in which the social actors face potential natural hazards and climate phenomena.“ (p. 16)
− Public infrastructure
− Housing conditions
− Nutrition
− Poverty and dependencies
− Economic capacity and income distribution
Vulnerability Coping
capacities
“ […] the capacities of societies and exposed elements (such as systems and institutions) to minimize the negative impact of natural hazards and climate change through direct action and resources.“ (p. 16)
− Government and authorities
− Disaster preparedness and early warning
− Medical services
− Social networks
− Material coverage Adaptive
capacities
“ […] capacities, measures and strategies that enable communities to change in order to address expected negative consequences of natural hazards and climate change.” (p.
17)
− Education and research
− Gender equity
− Environmental status/ ecosystem protection
− Adaption strategies
− Investments
Table 1: Definition and sub-categories of the components of the World Risk Index.
(all information in this table is retrieved from the World Risk Report 2011 made available by the Bündnis Entwicklung Hilft, 2011).
Figure 1: Composition of World Risk Index and its components.
(all information in this figure is retrieved from the World Risk Report 2011 made available by the Bündnis Entwicklung Hilft, 2011).
14 3.2. Data regarding the Different Markets
For each country, the market risk premium is examined as dependent variable. Thus, stock returns are an integral part for constructing the market risk premium. For the results to not suffer from survivorship bias (see Brown et al., 1992; Blake et al., 1993; Elton et al., 1996), companies still active as well as failed companies need to be included in the sample. In this study, survivorship bias might stem from stocks that disappear from the sample due to bankruptcy or delisting during the examination period. In both cases, the underlying reason is likely to be poor performance (Elton et al., 1996). Hence, neglecting failed companies might lead to an overstatement of results, because stocks that survive the sample period are more likely to have a stronger performance in general.
To download daily stock price return data for companies in the US, the Netherlands, and Japan, the data bases Thomson Reuters Datastream2 and EIKON3 were used. For each country, a list of active companies was generated in EIKON for which the variable Country of Headquarter (TR.HQCountryCode) needed to be equal to the respective country (the US, the Netherlands or Japan). Subsequently, the Company Common Name, the North American Industry Classification System (NAICS) Sector Code and Name, and the NAICS Subsector Code and Name were retrieved.
For the failed companies, the DEAD-Lists DEADNL, DEADJP, DEADUS1 and DEADUS2 were downloaded from Datastream. Consequently, the tickers were translated to the International Securities Identification Numbers (ISINs) to be able to download the static
2 Datastream is a data base on economic, financial, as well as legal data. As access to the data base is liable to cost and Thomson Reuters is competing with Bloomberg Inc., the data quality is regarded to be high. Further, Datastream is frequently used as a data base in highly published academic research (cf. Acharya et al., 2011;
Fauver et al., 2017). Datastream was accessed via designated desktop computers within the university library.
3 EIKON is a data base on economic, financial, as well as legal data. As access to the data base is liable to cost and Thomson Reuters is competing with Bloomberg Inc., the data quality is regarded to be high. Further, EIKON is frequently used as a data base in highly published academic research (cf. Du et al., 2018; Loh and Stulz, 2018) EIKON was accessed via designated desktop computers within the university library. Datastream and EIKON are interconnected because they are using the same underlying data library.
15 variables Country of Headquarter, Company Common Name, the NAICS Sector Code and Name, and the NAICS Subsector Code and Name. Finally, for all firms in the sample, the Total Return Index (X(RI)~U$) in US Dollars was downloaded from Datastream on a daily basis for the period of December 31, 2010 to January 02, 2017.
Time-series data on the market returns in the respective countries was downloaded from Datastream. To retrieve daily data on the market, the Total Return Index (X(RI)~U$) in US Dollars was downloaded for the Financial Times Stock Exchange (FTSE) market indices, FTSE Japan, FTSE Netherlands, and FTSE United States, in the period of December 31, 2010 to January 02, 2017. These indices encompass all securities traded in the respective market and thereby represent the return index of the total market. All three indices have the same base date from which the index is calculated.
The risk-free rate was downloaded from a data library provided by professor Kenneth R.
French4. In all datasets from the French data library, the 1-month Treasury bill of the United States is used as the risk-free rate, because it is not biased by illiquidity.
The data on the control variables, namely, Inflation (Consumer Price Index), Money growth M3, short-term interest rates, foreign exchange rates and forecasted growth in the real GDP were downloaded from the OECD Data library5. For each examined country, the data of the measures in each respective month was copied to a Microsoft Excel file which subsequently was saved as a txt-file. For the Dutch Money growth M3 measure the values for EA19 were used as this represents the money growth of the Euro Area, of which the Netherlands are a part
4 Via: https://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html. Professor French is a well- known and high-ranked scholar in the field of Finance. He holds and held positions at the MIT, Yale University, and the Tuck School of Business as well as senior positions in the practical field. Professor French together with Professor E. Fama are widely recognized for their three-factor model and accompanying research for which they won several prizes (cf. Fama, 2013; The American Finance Association, 2019; The Journal of Financial Economics, 2019). Therefore, the data in his data library is generally regarded as high-quality and up to date.
5 Via: https://data.oecd.org. The Organization of Economic Cooperation and Development (OECD) is a well- known intergovernmental organization aiming at improving lives through economic as well as political and social development. The OECD collects and analyses data themselves with the objective of being independent and evidence based. Due to its reputation, its role in the political and economic context, and its credibility, the data in the data library is regarded as high-quality and up to date and often used in academic research (cf. Nakamura et al., 2013; Fauver et al., 2017; Lei et al., 2018).
16 of. All measures are formatted as percentage numbers, meaning that a value of 5.89% is stated as 5.89. For the money growth measure, the year 2015 is the base year to which a value of 100%
is assigned, resulting in values of previous and subsequent years scaled around 100. Further, the foreign exchange rate for all currencies by default is the conversion rate to US Dollar, which is why for the United States, the measure for the foreign exchange rate is one for all years.
For the robustness checks the regional datasets containing the three Fama-French risk factors (see 1992; 1993) and the momentum factor (see Carhart, 1997) have been retrieved from the data library provided by professor Kenneth R. French4.
4. Methodology
4.1. Baseline Model
4.1.1. Derivation of the Baseline Model
To address the research question and to test the stated hypotheses, regressions based on the Fama and MacBeth (1973) methodology are performed. The Fama-MacBeth regression is a multistep process as pointed out in detail hereafter.
First, a time-series regression using an equilibrium model is performed. In the context of this study, the Capital Asset Pricing Model (CAPM) is used to estimate the betas of individual stocks in different countries over a certain period of time. The CAPM is chosen, because it is a consumption model. This line of thought is based on the assumption that consumption is expected to temporarily decrease in the event of a natural disaster taking place. Thus, the regression equation is represented by
𝑅𝑖,𝑡𝑒 = 𝛼𝑖+ 𝛽𝑖𝑅𝑚,𝑡𝑒 + 𝑢𝑖,𝑡, 𝑖 = 1, … , 𝑁; 𝑡 = 1, … , 𝑇 (1.1) where 𝑅𝑖,𝑡𝑒 denotes the excess return of an individual asset 𝑖 at trading day 𝑡, and 𝑅𝑚,𝑡𝑒 denotes the excess return of the market 𝑚 at trading day 𝑡. Furthermore, 𝛼𝑖 denotes the intercept and 𝛽𝑖
17 denotes the slope parameter, which is a measure of risk for the individual assets. Additionally, 𝑢𝑖,𝑡 represents the error term.
Following this time-series regression, a cross-sectional regression using the averages of individual stock returns per calendar month 𝑡 is performed by applying
𝑅̅𝑖 = 𝜆0+ 𝜆1𝛽̂𝑖 + 𝑣𝑖, 𝑖 = 1, … , 𝑁 (2.1)
where 𝑅̅𝑖 denotes the average return of an individual asset 𝑖. 𝜆0 represents the constant and can be interpreted as the risk-free rate, whereas 𝜆1 is the slope parameter which represents the market risk premium. 𝛽̂𝑖 should be equal to the actual beta of an individual asset 𝑖, and 𝑣𝑖 denotes the residuals of the individual asset 𝑖.
To examine the factors that influence the market risk premium ̂1, an additional regression is performed (see Equation (3.1)). In this regression, the market risk premium (𝜆̂1,𝑡) is the dependent variable and the overall score of the World Risk Index (𝑊𝑅𝐼𝑗,𝑡) represents the independent variable. The regression equation is represented by
𝜆̂1,𝑡 = 𝛿1𝑊𝑅𝐼𝑗,𝑡+ 𝛿2𝐼𝑁𝐹𝐿𝑗,𝑡+ 𝛿3𝑀𝐺3𝑗,𝑡+ 𝛿4𝐼𝑁𝑇𝑅𝑗,𝑡+ 𝛿5𝐹𝑋𝑗,𝑡+ 𝛿6𝐺𝐷𝑃𝑔𝑗,𝑡+ 𝜀𝑗,𝑡,
𝑗 = 1, … , 𝐽; 𝑡 = 201101, … , 201612
(3.1)
where 𝑊𝑅𝐼𝑗,𝑡 denotes the overall World Risk Index score calculated in the World Risk Report (Bündnis Entwicklung Hilft, 2011) of country 𝑗 in calendar month 𝑡.
Following the inherent relationship of risk and return, a positive sign of 𝛿1 is expected, because a higher natural disaster risk should lead to an increase in the demanded compensation of risk-averse investors and therefore, the market risk premium should increase.
To build a valid model and to prevent endogeneity problems caused by omitted variables, control variables are included in Equation (3.1). In this study, macroeconomic variables that exert influence on the market risk premium are used as control variables. Prior literature (cf.
Bodie, 1976; Fama, 1981; Geske and Roll, 1983; Benaković and Posedel, 2010) demonstrates
18 the significant impact of inflation and money growth on aggregated stock returns. Therefore, measures for inflation (INFL) and money growth (MG3) are included in the analysis. Money growth is measured by the broader money measure M3, because the M1 measure has less explanatory power (Fama, 1981). The M1 measure only includes the currency, whereas the broader measure M3 additionally includes deposits, money market funds, and debt securities.
In addition to inflation, Obadire (2018) determines that the interest rate and foreign exchange rates are factors influencing the market risk premium. Consequently, the variables INTR representing the short-term interest rates and FX representing the foreign exchange rates to US Dollar are included in the regression. Additionally, Tahmidi et al. (2011) highlight the significant influence of the forecasted growth in real GDP, GDPg, on the market risk premium.
Hence, the variable GDPg is included in the regression, too.
To examine whether 𝛿1 has a statistically significant influence on the market risk premium, a t-test has to be implemented. If the absolute value of the T-statistic is larger than the critical value, the alternative hypothesis 𝐻1: 𝛿1 0 should be accepted and the null hypothesis 𝐻0: 𝛿1 = 0 should be rejected. In case of accepting the alternative hypothesis 𝐻1, the natural disaster risk, as proxied by the World Risk Report, has a statistically significant influence on the market risk premium at the 𝛼-level. The magnitude and direction of the influence have to be examined in regard of the sign and value of parameter 𝛿1.
4.1.2. Limitations of the Baseline Model
To ensure the validity of the results, the weaknesses of the proposed research design need to be considered, too. First, and foremost, endogeneity problems may arise. Although a simultaneous relation between the dependent variable (i.e., market risk premium) and one independent variable (i.e., exposure, which is fully exogeneous) seems unlikely, there might be a simultaneous relation between the market risk premium and vulnerability of a country. On the one hand, vulnerability incorporates, among other factors, the lack of both coping activities and
19 adaptive capabilities. These characteristics can be observed to a stronger degree for less developed countries (see Bündnis Entwicklung Hilft, 2011). On the other hand, research shows that the risk premia for emerging markets are significantly higher than for developed economies (see Salomons and Grootveld, 2003). Keeping these two observations in mind, spurious results may be observed and, hence, lead to biased conclusions in terms of coefficients of the components being too high and/or significant, even if the effect itself is not. To address this potential issue, the country sample only includes advanced economies (International Monetary Fund, 2016), namely, the United States, the Netherlands, and Japan.
The omittance of variables might be another source from which endogeneity problems might arise. The dependent variable, the market risk premium, incorporates all systematic risks in the market (see Section 2.1) and, therefore, is influenced by many, partly unobserved, factors.
Due to the nature of this influence of unobservable determinants on the market risk premium, it is not feasible to include all determinants in the analysis to identify each effect separately.
Hence, in the context of this thesis, omitted variables are likely. As outlined during the derivation of the baseline model, the controls used in this analysis are grounded in literature (see Bodie, 1976; Fama, 1981; Geske and Roll, 1983; Benaković and Posedel, 2010; Tahmidi et al., 2011; Obadire, 2018) to ensure their relevance. However, the selection of the controls was constrained by data availability and meaningfulness, as in the case of the foreign exchange rate to US Dollars being equal to one for the US. Therefore, omitted variable bias in this analysis cannot be ruled out entirely.
4.2. Model regarding Coping Abilities
The baseline model outlined in Section 4.1 needs to be modified in order to be able to test Hypothesis H2. As Hypothesis 2 examines the influence of a country’s vulnerability on its market risk premium, Equation (3.1) needs to be altered as follows
20 𝜆1,𝑡 = 𝛿1𝑒𝑥𝑝𝑜𝑠𝑢𝑟𝑒𝑗,𝑡+ 𝛿2𝑣𝑢𝑙𝑛𝑒𝑟𝑎𝑏𝑖𝑙𝑖𝑡𝑦𝑗,𝑡+ 𝛿3𝐼𝑁𝐹𝐿𝑗,𝑡+ 𝛿4𝑀𝐺3𝑗,𝑡+
𝛿5𝐼𝑁𝑇𝑅𝑗,𝑡+ 𝛿6𝐹𝑋𝑗,𝑡+ 𝛿7𝐺𝐷𝑃𝑔𝑗,𝑡+ 𝜀𝑗,𝑡, 𝑗 = 1, … , 𝐽; 𝑡 = 201101, … , 201612
(3.2)
In this equation, the second explanatory variable 𝑣𝑢𝑙𝑛𝑒𝑟𝑎𝑏𝑖𝑙𝑖𝑡𝑦𝑗,𝑡 is added to explore the driver of the initial relationship between overall natural disaster risk and the market risk premium. The variable 𝑣𝑢𝑙𝑛𝑒𝑟𝑎𝑏𝑖𝑙𝑖𝑡𝑦𝑗,𝑡 is proxied by the vulnerability score as presented in the World Risk Reports, in which it is composed by the equally-weighted average of the single scores for susceptibility, lack of coping capacities and lack of adapting capacities (cf. Figure 1, p.13). The coefficients 𝛿1 and 𝛿2 now indicate the influence of exposure and vulnerability on the market risk premium and subsequently, the coefficients of the control variables are denoted with 𝛿3 to 𝛿7. The dependent variable, 𝜆1,𝑡 remains unchanged.
The fact that the World Risk Index is calculated by multiplying the two variables exposure and vulnerability, the coefficients of these variables should have the same sign as the coefficient of the variable 𝑊𝑅𝐼𝑗,𝑡 as used in Equation (3.1), namely a positive one.
To establish which component of 𝑣𝑢𝑙𝑛𝑒𝑟𝑎𝑏𝑖𝑙𝑖𝑡𝑦𝑗,𝑡 drives the results, the analysis is refined further by applying
𝜆1,𝑡 = 𝛿1𝑒𝑥𝑝𝑜𝑠𝑢𝑟𝑒𝑗,𝑡+ 𝛿2𝑠𝑢𝑠𝑐𝑒𝑝𝑡𝑖𝑏𝑖𝑙𝑖𝑡𝑦𝑗,𝑡+
𝛿3𝑙𝑎𝑐𝑘 𝑜𝑓 𝑎𝑏𝑖𝑙𝑖𝑡𝑖𝑒𝑠 𝑡𝑜 𝑐𝑜𝑝𝑒𝑗,𝑡+ 𝛿4𝑙𝑎𝑐𝑘 𝑜𝑓 𝑎𝑏𝑖𝑙𝑡𝑖𝑒𝑠 𝑡𝑜 𝑎𝑑𝑎𝑝𝑡𝑗,𝑡+ 𝛿5𝐼𝑁𝐹𝐿𝑗,𝑡+ 𝛿6𝑀𝐺3𝑗,𝑡+ 𝛿7𝐼𝑁𝑇𝑅𝑗,𝑡+ 𝛿8𝐹𝑋𝑗,𝑡+ 𝛿9𝐺𝐷𝑃𝑔𝑗,𝑡+ 𝜀𝑗,𝑡, 𝑗 = 1, … , 𝐽; 𝑡 = 201101, … , 201612
(3.3)
In this equation, the variable 𝑣𝑢𝑙𝑛𝑒𝑟𝑎𝑏𝑖𝑙𝑖𝑡𝑦𝑗,𝑡 is replaced by variables representing its components. The variables 𝑠𝑢𝑠𝑐𝑒𝑝𝑡𝑎𝑏𝑖𝑙𝑖𝑡𝑦𝑗,𝑡, 𝑙𝑎𝑐𝑘 𝑜𝑓 𝑎𝑏𝑖𝑙𝑖𝑡𝑖𝑒𝑠 𝑡𝑜 𝑐𝑜𝑝𝑒𝑗,𝑡 and 𝑙𝑎𝑐𝑘 𝑜𝑓 𝑎𝑏𝑖𝑙𝑡𝑖𝑒𝑠 𝑡𝑜 𝑎𝑑𝑎𝑝𝑡𝑗,𝑡 are proxied by their components’ scores as presented in the World Risk Reports respectively. Therefore, the coefficients 𝛿2, 𝛿3 and 𝛿4 now indicate the influence
21 of the single components on the market risk premium. Consequently, the coefficients of the control variables are denoted with 𝛿5 to 𝛿9.
Following the same line of reasoning as for the World Risk Index and its components, the coefficients 𝛿2, 𝛿3 and 𝛿4 are expected to have the same, positive, sign as the coefficient 𝛿2 in Equation (3.2) (i.e., the coefficient of the variable 𝑣𝑢𝑙𝑛𝑒𝑟𝑎𝑏𝑖𝑙𝑖𝑡𝑦𝑗,𝑡).
The necessity of this further refinement is due to the fact that the variable 𝑣𝑢𝑙𝑛𝑒𝑟𝑎𝑏𝑖𝑙𝑖𝑡𝑦𝑗,𝑡 is an aggregate measure of its components. To be able to determine which of the three components is the driving factor in the influence on the market risk premium yields a higher meaningfulness and, hence, enables to determine more precise implications regarding the overall topic of natural disaster risk in a financial context.
In either case, the t-test initially only performed for 𝛿1 also needs to be performed for 𝛿2 in respect of Equation (3.2) and for 𝛿2, 𝛿3 and 𝛿4 respectively in case of Equation (3.3).
4.3. Model regarding Industry Specifications
To address Hypothesis H3, the industry-specific risk premium needs to be estimated per industry 𝑖𝑛𝑑 in country 𝑗 for each calendar month 𝑡 to retrieve 𝜆1,𝑡,𝑖𝑛𝑑(𝑗). To do so, individual stocks in each country are assigned to sub-samples according to the NAICS classification in the original dataset. The NAICS sector codes and names of the examined industries are 11 Agriculture, Forestry, Fishing and Hunting; 21 Mining; 22 Utilities; 23 Construction; 31-33 Manufacturing; 42 Wholesale Trade; 44-45 Retail Trade; 48-49 Transportation and Warehousing; 51 Information; 52 Finance and Insurance; 53 Real Estate Rental and Leasing;
54 Professional, Scientific, and Technical Services; 55 Management of Companies and Enterprises; 56 Administrative and Support and Waste Management and Remediation Services;
61 Educational Services; 62 Health Care and Social Assistance; 71 Arts, Entertainment, and Recreation; 72 Accommodation and Food Services; and 81 Other Services (excl. Public
22 Administration). Based on these, respective sub-samples per sector and countries are composed and the regression analyses as outlined in Section 4.2 are performed.
The impact of natural disaster risk on the industry-specific risk premium is expected to vary between industries, because industries are differently impacted by disasters (see Guidry and Pruitt, 2012; Loayza et al., 2012; Ulubaşoğlu et al., 2019). In general, there are three groups of industries that can be identified: (1) industries strongly negatively affected, (2) industries almost not affected, (3) industries possibly positively affected. One representative of category (1) is the Agricultural industry. The Agricultural industry is expected to be directly, negatively impacted by natural disasters, because especially floods, droughts and earthquakes are able to destruct agricultural assets on a large scale without an option to recover or reuse these assets.
Similarly affected industries might be Manufacturing, Transportation or Real Estate Rental and Leasing, because they are heavily reliant on tangible assets, such as production sites, machinery, infrastructure and real estate. The destruction of these assets is expected to lead to a decrease in output and hence, revenue, although some of these assets may be replaceable. Due to the negative impact of natural disasters, natural disaster risk should lead to a higher demand for an additional compensation for investing in companies in these industries than in other industries.
Hence, the industry-specific risk premia should be positively affected by natural disaster risk, indicating that the regression coefficients of the natural disaster risk measures should be positive.
In industries not heavily reliant on tangible assets, the impact of natural disasters risk might be rather small. Such industries are mainly related to service-related business activities and, therefore, include Information, Professional Services, Management of Companies, Administrative Services, Educational Services and Other Services. Additionally, Arts and Entertainment, Accommodation and Food Services, Mining and Utilities should not be severely affected by natural disasters as the products and services provided by these industries are demanded for essential survival and/or irrespective of whether there is a natural disaster or not.
23 Therefore, all these industries should almost not be affected and, hence, belong to group (2). It is expected that the regression coefficients of the natural disaster risk measures are close to zero.
Extreme weather events and natural disasters are likely to destroy a wide range of assets, including tangible assets, such as housing, office space and manufacturing sites, but also infrastructure, such as heavy and civil engineering construction (e.g., bridges, buildings or streets). As these goods can be rebuilt, the destruction of these assets is expected to lead to an increased demand for construction services subsequent to a natural disaster. This increased demand might offset the natural disaster risk in the industry. A similar example are Health Services, which often are highly demanded after a natural disaster. Further, Wholesale as well as Retail Trade should experience an increased demand due to the destruction, especially of private goods. Thus, the industry-specific risk should not increase. Subsequently, the industry- specific risk premia for these industries belonging to group (3) are expected to be negatively affected by natural disaster risk. Depending on the magnitude of the demand increase the regression coefficients of the natural disaster risk measures should be negative.
For the Financial and Insurance industry, the categorization is not entirely clear, because the impact of a natural disaster and, hence, natural disaster risk, depends on the timing of the disaster (cf. Jenkins, 2019). On the one hand, if natural disaster risk is perceived by individuals and firms, the likelihood that these individuals and firms seek insurance is high. This represents an advantage for insurance companies because the insurance premiums generate revenues. On the other hand, in case of a natural disaster, large liabilities are due to the insured parties which have to be paid by the insurer. This, in turn, reduces the revenues and leads to a weaker performance. Additionally, after a natural disaster, insurance premiums often increase (Reuters, 2019) which leads to a larger insurance gap, because a larger number of individuals (and companies) might not be able to fund these larger payments. This, subsequently, decreases the revenues within the Insurance industry. Therefore, from a pure risk perspective, the industry-
24 specific risk premium for the Finance and Insurance industry should decrease in case of increased natural disaster risk (i.e., negative coefficients) due to an increased opportunity for business resulting from a higherrisk. However, when anticipating the actual disaster, this expectation might turn towards an increased industry-specific risk premium in case of increased natural disaster risk (i.e., positive coefficients).
5. Empirical Results and their Discussion
5.1. Descriptive Statistics and Correlation Matrices providing first Insights 5.1.1. Descriptive Statistics
For the three examined countries the sizes of the samples vary according to the number of public firms available in the data bases that fulfilled the criteria as described in the data collection (see Section 3.2). After the data preparation, the final samples comprise 5,219,302 firm-level observations for Japan, 487,976 for the Netherlands, and 14,604,479 for the United States. The variation in the sample sizes is not surprising as the analysed markets and economies differ in their size, too. For the largest financial market by size and volume, the US, the highest number of observations is included in the sample. In line with this, the sample of the smallest market, the Netherlands, comprises the least firm-level observations.
Table 2, Table 4 and Table 6 (pp. 26, 27, 28) report the descriptive statistics of the variables used in the subsequent analyses for Japan, the Netherlands and the US respectively.
The Market Risk Premium is reported to have a rather small magnitude and standard deviation especially in Japan and the Netherlands. In the US, the larger range of values (from - 3.879 to 10.139) in combination with the median remaining around 0 indicates outliers at the borders of the range. The variable World Risk Index indicates the overall risk of natural disasters with a higher value suggesting a higher risk. The World Risk Index is the lowest for the United States and the highest for Japan, leaving the Netherlands in between. This indicates that the US,
25 overall, is at the lowest risk of extreme natural weather events turning into natural disasters. To put the mean values of 13.08 (Japan), 8.29 (Netherlands) and 3.868 (US) into context, the ranks of the countries can be taken into account. Japan, on average, is situated in the top 10% of the ranking in most years, indicating a high overall risk of natural disasters. The Netherlands are positioned around the 50th place (out of 173 places) in all years, therefore, belonging to the top 30% of the countries most at risk. The United States belong to the bottom quartile of the ranking, with positions around the 130th place. These rankings provide an overview on how much at risk the single countries are in a global context.
A similar picture emerges, when looking at exposure. Japan experiences the highest value of exposure, whereas the US has to lowest exposure to extreme weather events. With a value of 45.91 for five out of six years, Japan ranks as the 4th most exposed country in the world, mainly due to possible earthquakes and/or tsunamis as happened in 2011 (International Atomic Energy Agency, 2015). The Netherlands are experiencing the second highest exposure of the three examined countries due to the high risk of floods and the continuous threat of rising sea levels. This phenomenon puts many people at risk, because 60% of people live in areas below sea level (United Nations University, 2016) and, therefore, poses a high threat to the country in general. Of the three countries, the US have the lowest exposure which mainly results from a mixture of phenomena. Hurricanes and floods are a threat to certain parts of the US (National Weather Service, 2019a; National Weather Service, 2019b). Further, wildfires are common in some areas like California (see Californian Department of Forestry and Fire Protection, 2019) and lower impact weather events like tornados and blizzards (see BBC, 2014; Berardelli, 2019) might be experienced throughout the country. When looking at the composition of exposure, these lower impact events (i.e., tornados and blizzards) are not included in the measure, and neither are wildfires, because these types of events are not “responsible for most of the human casualties and material damage“ (Bündnis Entwicklung Hilft, 2011, p. 18). This leaves the US with the lowest exposure of all three countries, possibly contrary to prior expectations.
