Climatic Change and Financial Stability: Natural Disaster Impacts on Global Stock Markets

This paper aims to provide a comprehensive study of the impacts of worldwide climatic change and consequent natural disasters on international stock markets. By means of a suited event study methodology, we investigate the eﬀects of biological, climatological, geophysical, hydrological and metereological disasters occurred in 104 countries across the world on 27 global stock market indexes over the period 8 February 2001 to 31 December 2019. We ﬁnd diverse stock market responses to natural hazard shocks depending on the type of event under consideration, as well as on the location in which the event has occurred. We discover that climatological and biological calamities are the disaster types which induce the most extreme reactions of international ﬁnancial markets, followed by geophysical ones. Furthermore, the examined stock indexes are, on average, considerably responsive to shocks occurring in countries belonging to the European continent, which, overall, tend to aﬀect in a negative way their performances. Finally, our empirical investigation sheds light on the diversiﬁcation opportunities arising from the mitigation of natural catastrophe risks, by providing evidence on the sensitivity of stock indexes to disaster-speciﬁc and country-speciﬁc natural hazards. A natural disaster risk hedging strategy highlights the diversiﬁcation opportunities arising from the mitigation of natural catastrophe risks, by providing evidence on the proﬁtability of trading stock indexes hedging for speciﬁc natural hazard sources, and particularly climatological and biological ones.

costs that countries across the world need to bear. As a matter of fact, direct 26 losses from natural disasters given as a share of Gross Domestic Product (GDP) are 27 estimated to range from 0.12% to 0.5% of global GDP over the period 1990-2017 2 . 28 Natural catastrophes can be regarded as non-financial, exogenous shocks to the 29 economy -see e.g. Skidmore and Toya (2002), Ramcharan (2007), Yang (2008), 30 Raddatz (2009), Mahajan and Yang (2020). Besides affecting several macroeco-31 nomic indicators, they have also direct impacts on domestic financial markets, as 32 well as they exert effects which might reverberate across financial markets of vari- We contribute to the extant literature regarding the impact of natural disasters 54 on international financial markets in several ways. Differently from most of the 55 earlier research, we do not limit our analysis to domestic natural catastrophes: we 56 analyze the effects of natural hazards occurred during the last two decades across 57 the world on price changes of major and geographic widespread aggregate stock 58 market indexes. To this aim, we tailor our event study methodology to take into ac-59 count for specific economic and financial dimensions of each country's corresponding 60 financial index, besides controlling for specific time series features. Additionally, we 61 do not only examine the impact of some specific sub-group of natural hazards (e.g.  The remainder of this paper proceeds as follows. Section 2 provides a literature 78 review on the topic and methodologies here studied. Section 3 gives details on 79 the methodology we employ in order to conduct the event study. In Section 4 we 80 5 illustrate the data and our preliminary analysis. In Section 5 we present and discuss 81 our empirical outcomes. Section 6 illustrates the empirical outcome of our proposed 82 natural disaster risk hedging strategy. Section 7 concludes.  find that no individual country stock market is affected by the contagion effect, but 100 that the foreign exchange markets of some countries suffered from it. ing that the negative sentiment due to bad mood and anxiety affects the decision-135 making process of market participants, which in turn influence asset pricing. Ka-136 planski and Levy (2010), for instance, examine the impact of aviation disasters on 137 stock prices throughout an event study. They find evidence of a significant nega-    (Binder, 1985).

174
Let us consider the continuously compounded returns time series R i,t , computed 175 as: where P i,t and P i,t−1 are the prices of the generic market index i at time t and t − 1, respectively. The ARs can be parametrized by means of the inclusion of an 178 event-day dummy variable in the market model, as follows: where α i and β i stand for the market alpha and beta, respectively, and R m,t repre-  whereas ǫ i,t is a zero-mean error term.

187
In order to quantify the overall reaction in financial indexes following the natural 188 disaster events, ARs can be aggregated after the SUR estimation to derive the 189 cumulative abnormal return (CAR) over the event window [t 0 , t w ] for each financial The fundamental model presented in Equation (2)   consequence, the set of exogenous control variables in our empirical analysis is given

217
Our aim is to discover both disaster-specific and location-specific effects on world-218 wide financial indexes. Thus, we design our regression analysis in a twofold way.

219
Firstly, we consider the impact on the considered financial indexes, of all groups of 220 events (i.e. biological, climatological, geophysical, hydrological and metereological), 221 regardless of the country in which the event has occurred. In this case, the param-222 eter γ i,t represents the AR on stock index i at time t due to a particular category 223 of natural hazard. Secondly, we assess the impact on the sampled financial indexes 224 of natural disasters occurred in one specific country, regardless of the type of event.

225
In this case, the parameter γ i,t represents the AR on market index i at time t due 226 to events hitting a particular country. In order to conduct our empirical analysis, we combine different sources of data.

229
Firstly, we analyze the international Emergency Events Database (EM-DAT), con- substances (e.g. venom, mold) or vector-borne diseases that they may carry.

249
In Figure 1 we illustrate the geographic distribution of worldwide natural disas-    (related to the MOEX Russia index), which is evidence of a generally leptokurtic 298 behaviour with respect to a benchmark normal distribution. 5 Empirical results and discussion 300 We present our empirical results as follows. In the first Subsection, we examine the 301 impact of each type of natural disaster on the performances of each market index.

302
In the second Subsection, we illustrate how ARs vary according to the geographical 303 location of the natural hazards.  The Great East Japan Earthquake of 2011, besides others, damaged many chemical installations, including a refinery which was inundated by the tsunami originating a structural damage. Storage tanks containing sulfur, asphalt and gasoline caught fire. Source: Chemical releases caused by natural hazard events and disasters, WHO (2018): https://reliefweb.int/report/world/ chemical-releases-caused-natural-hazard-events-and-disasters-information-public-health 8 The combination of storms and high winds occurred during hurricane Katrina generated oil spills from refineries, releases of diesel fuel from tanks, waste sites and abandoned vehicles, as well as remobilization of soil contaminants. Source: Chemical releases caused by natural hazard events and disasters, WHO (2018): https://reliefweb.int/report/world/ chemical-releases-caused-natural-hazard-events-and-disasters-information-public-health 9 Estimates of the total damages (USD) caused by natural catastrophes expressed are those according to the EM-DAT database.    The majority of geophysical events impact financial markets in a negative way.

379
The largest significant negative effect is that on the Hong Kong Hang Seng index.

380
China is indeed the country which suffered the largest number of geophysical hazards 381 during the considered period, many of which caused devastating economic impacts. Evidence also suggests that the Nasdaq Composite and Nasdaq 100 indexes react 406 positively when meteorological calamities occur. Hence, in order to mitigate meteo-407 rological risks, it seems convenient to invest in technological sector indexes such as 408 the Nasdaq Composite or Nasdaq 100, whose stock composition and geographical 409 coverage enhance resilience to shocks arising from meteorological hazards.  10 See, for instance, the 2009 exceptional winter storm over northern Iberia and southern France -the so called Klaus cyclone -which caused massive damages to properties and major forests in the Spanish country, and the European heat wave of 2003, which affected a significant portion of western Europe, with Spain counting more than 15,000 deaths.

23
In this setting we obtain, for each considered market index, an estimate of the ARs 417 caused by the occurrence of natural calamities in each world country. In order to 418 provide a comprehensive overview of the AR dynamics across market indexes and 419 countries, we present aggregate results by continents in which events have occurred. 420 Particularly, we consider the impact of natural disaster shocks occurred in Europe, 421 America and Asia. This is illustrated in Figure 6, where we show the estimated av-  Figure 6: Estimated average ARs from natural disasters in Europe, America and Asia. The figure shows the estimated average γ i,t associated to natural disasters occurring in European, American and Asian countries by market index. We consider the average of statistically significant effects, namely those coefficients reporting a p-value which is less than 1%.

25
The magnitude of the average CAR coefficients associated to natural disasters 428 occurring in world continents shows that market indexes respond heterogeneously 429 to natural shocks depending upon the countries in which they take place. Indeed,  The trading strategy is back-tested using a walk forward approach. We opt for an   To comprehensively measure the actual risk-return profiles of our set of top-580 bottom portfolios, we also compute Sharpe ratios.     Oceanian countries by market index. We consider the average of statistically significant effects, namely those coefficients reporting a p-value which is less than 1%.