The impact of natural disasters on stock markets: evidence from Japan and the US.
Wang, Lin ; Kutan, Ali M.
INTRODUCTION
Japan and the United States (US) are among the countries that have
suffered extensively from natural disasters, especially earthquakes,
tsunamis, and volcanoes, due to their geographic location. Japan lies
right at the junction of the Pacific plate and the Eurasian plate while
the US is adjacent to the Pacific plate. The movement of these plates
often causes earthquakes and volcanoes, many of which are associated
with tsunamis. A 9.0-magnitude earthquake struck the Pacific coastal
areas of Japan on 11 March 2011, churning up a devastating tsunami. It
was the most powerful quake on record to hit Japan. As of 25 April 2011,
the death toll reached 14,358 while 11,889 were listed as missing. (1)
The nuclear crisis that resulted from the earthquake made the situation
even worse. The financial loss is expected to reach 309 billion US
dollars, the largest loss from a natural disaster in the world. (2) The
US also suffered from earthquakes, tsunamis, and hurricanes. For
example, according to the National Weather Service and the Insurance
Information Institute, damage of more than $501.1 billion and a death
toll of 22,240 resulted from the top 10 US natural disasters during
1980-2010. (3)
Another reason that makes Japan and the US of interest is that the
two countries are among the largest economies in the world. The Japanese
Nikkei 225 and the Standard and Poor's 500 are broadly cited stock
indices, and the Japanese Yen and the US dollar are the most frequently
used currencies in international trade as well as in financial markets.
If the frequent natural disasters have a significant impact on the
domestic financial markets, such as on the insurance companies and the
composite stock index, there might be contagious effect on other
financial markets around the world. Researchers have shown that natural
disasters have an impact on the insurance industry (Yamori and
Kobayashi, 2002; Shelor et al., 1992). Also, there were fluctuations in
the Nikkei 225 and the Standard and Poor's 500 stock indexes after
the 11 March earthquake and tsunami in 2011. Since the composite stock
index represents a well-diversified portfolio, it is a better measure of
the overall performance of the financial market than a single asset,
such as crude oil or gold, which usually shows higher volatility in
return due to lack of diversification. According to the Capital Asset
Pricing Model, any risk that could potentially be reduced by
diversification should not yield extra return beyond the market return.
What really matters is the systematic risk that cannot be diversified
away (Ross et al., 2010). In this case, if there are any changes in the
stock market return that are due to natural disasters, special attention
to the impact of natural disasters will be necessary since these impacts
could not be effectively absorbed by the market and may lead to
significant wealth losses when they happen again in the future. However,
some studies summarized in the review section have reported some gains
in stock returns in the insurance sector due to natural disasters and
they explained this based on the so-called 'gaining from loss
hypothesis', which suggests that investors tend to demand more
insurance coverage during times of natural disasters to maximize their
protection and hence insurance sector profits increase yielding higher
stock returns in this sector.
The objective of this paper is to investigate whether there is any
undiversifiable impact of natural disasters on the insurance sector
stock market return as well as on the composite stock market return in
Japan and the US. Such impact is referred to as the wealth effect. This
paper also examines the risk effect, which is the impact of natural
disasters on stock market volatility.
The paper is divided into seven sections. Following this
introduction, the next section reviews the existing literature. The
subsequent section discusses the contribution of this paper. The
methodology employed in this study is explained in the latter section.
The following section describes data and provides the descriptive
statistics. The penultimate section discusses the estimation results and
findings. Finally, conclusion and implications are provided in the final
section.
LITERATURE REVIEW
Although several studies examine the impact of natural disasters
from an economic perspective, most of them focus on the property
liability insurance industry. However, there are no consistent
conclusion on whether the stock prices of insurance companies drop or
rise after severe natural disasters.
Yamori and Kobayashi (2002) investigate the impact of the Japanese
Hanshin-Awaji earthquake, which hit the Tokyo metropolitan area in 1995,
on the stock prices of Japanese insurance firms. This earthquake cost
the insurance companies approximately 77 billion yen, the largest
payment since the Japanese earthquake insurance system was established.
Their study is the first to test the gaining from loss hypothesis
outside the US. They use an event study methodology based on Ordinary
Least Squares (OLS), and, by creating a portfolio of 13 insurance
companies, they calculate the daily abnormal returns of the portfolio
from day 0 to day 9 after the earthquake happens. They get a
significantly negative abnormal return on day 0, which is the day the
earthquake. The results are quite consistent based on pre- and
post-period estimations. They reject the hypothesis that insurance
companies can benefit from increased demand for their products after a
natural disaster, contrary to some studies for the US insurance sector.
The authors do not provide possible explanations of such difference.
This study employs a small number of companies included in the
portfolio, possibly leading to biased estimates.
Lamb (1995) also reports evidence of a negative stock price
response on property liability insurers after Hurricane Andrew hit
Florida and Louisiana in 1992. The sample consists of 37 publicly held
property liability insurance companies. Daily stock returns are used to
calculate abnormal returns from 10 days before the hurricane to 30 days
after it. The results reveal negative stock price responses of property
liability insurance companies that are exposed to the losses. However,
no significant stock price response is observed for insurance companies
without loss exposure in the two states. The estimation of stock betas
is only based on the pre-event period without testing stability of betas
before and after the hurricane. Stock betas are very likely to change
due to volatility change caused by the natural disaster.
Shelor et al. (1992) examine the 'gaining from loss'
hypothesis based on the 1989 California earthquake in the US and reach
different conclusions. Their study also adopts an event study
methodology, but uses Generalized Least Squares and Non-generalized
Least Squares for estimation. They create two portfolios one with
property liability insurers and the other with multiple product
insurers, with sample sizes of 47 and 32, respectively. Daily stock
returns are used to calculate abnormal returns from day 0 to day 15
after the earthquake. They test risk stability for the pre-event period
and post-event period and do not observe any change in stock betas.
Their result supports the gaining from loss hypothesis, because they
find evidence of significant increase in stock value of both portfolios
after the earthquake. From a cross- sectional analysis, they also
conclude that stock value of California insurance companies had a
smaller increase than that of companies that did not write insurance in
California because the positive effect of California insurance companies
is partially offset by the loss payments.
Worthington and Valadkhani (2004) examine the impact of natural
disasters on the Australian stock market by using intervention analysis
based on an Autoregressive Moving Average (ARMA) model. The data
employed are daily stock market returns covering the period of 31
December 1982 to 1 January 2002. They conclude that natural disasters,
especially cyclones and bushfires, have an impact on Australian stock
market and the net effects can be positive and/or negative depending on
the adjustment in the days that follow the disasters.
Another study by Worthington (2008) adopts the Generalized
Autoregressive Conditional Heteroskedasticity (GARCH) and the
GARCH-in-the-mean model based on the Australian stock market. The data
are daily stock market returns covering the period of 1 January 1980 to
30 June 2003. The results indicate that there is no significant impact
of natural disasters on Australian stock market returns. The shocks do
not create systematic risk in the market, and they are diversified away
at the market level.
CONTRIBUTION
First of all, this paper contributes to the literature by extending
the topic geographically to Japan and the US, and by investigating the
impact of disasters on both the insurance sector and the composite stock
market. Second, compared with previous studies, most of which use OLS,
this study employs the GARCH model, which is the same methodology that
Worthington (2008) uses for Australian stock market. However, this paper
improves the analysis by including necessary control variables that
Worthington (2008) omits, for example, foreign stock returns, the
exchange rate, and the interest rate, etc. Moreover, this study examines
the impact of natural disasters not only in the mean equation, but also
in the conditional variance equation, which is used as a proxy for risk.
METHODOLOGY
GARCH models are employed in this study. First of all, GARCH models
allow time-varying volatility while OLS assumes constant volatility. As
shown in Figure 1, it is clear that GARCH model is a better model than
OLS because the amplitude of the series varies over time. Also, compared
with the ARCH model, GARCH model captures the impact of all the past
shocks and it is more parsimonious because it uses past volatility
rather than all lags of shocks. TARCH and E-GARCH models both take into
account the asymmetric term, which measures the different impact of bad
news and good news on volatility (Enders, 2010). In this study, the
E-GARCH (1, 1) model is selected since it has the highest maximum
log-likelihood value. (4)
The model of stock market returns is described below:
Mean (conditional return) equation capturing wealth effects: Japan:
[Ret_jp.sub.t] = [[alpha].sub.0] + [[alpha].sub.1] [Ret_us.sub.t-1]
+ [[alpha].sub.2] [Chg_IR_jp.sub.t-1] + [[alpha].sub.3]
[Ret_EXR.sub.t-1] + [5.summation over (k=0)] [[alpha].sub.4+k]
ND_[jp.sub.t-k] + [[epsilon].sub.t] (1)
US:
[Ret_us.sub.t] = [[alpha].sub.0] + [[alpha].sub.1] [Ret_jp.sub.t] +
[[alpha].sub.2][Chg_IR_us.sub.t-1] + [[alpha].sub.3][Ret_EXR.sub.t-1] +
[5.summation over (k=0)] [[alpha].sub.4+k] [ND_us.sub.t-k] +
[[epsilon].sub.t] (2)
[FIGURE 1 OMITTED]
The dependent variables (Ret_US and Ret_Jp) are Japan and US
(insurance sector and composite) stock returns. The stock market return
of the foreign country, change in interest rate in the corresponding
local market (Chg_IR), and the US Dollar-Japanese yen exchange rate
return (Ret_EXR) are used as independent variables. Since major stock
markets around the world affect each other, and big events have an
impact on worldwide stock markets simultaneously, the performance of the
US stock market should be highly correlated with the performance of the
Japanese stock market. Also, changes in the interest rate may influence
the local stock market. Appreciation and depreciation of the domestic
currency may also bring about some changes in stock returns. Also,
because the stock market index, the interest rate, and the exchange rate
are financial market variables and may be affected simultaneously by
domestic or global events, they are likely endogenous. Therefore, to
avoid this problem, lag 1 terms of change in interest rate and return on
exchange rate are used as control variables.
Current and five lags of natural disasters are included in the mean
equation to capture their immediate and dynamic wealth effects on stock
returns, respectively. Natural disasters include four different
categories, namely, earthquakes, tsunamis, volcano eruptions and
tropical cyclones, and each enters into the regressions separately as
(0,1) dummy variables where 1 indicates that a natural disaster takes
place at a given date.
The model of the conditional volatility of stock market returns is
described below:
Conditional variance equation measuring risk effects: Japan:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (3)
US:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (4)
In the conditional variance equations, [[beta].sub.0] captures the
constant or unconditional variance, while [[beta].sub.1] measures the
impact of last period's shock on the conditional variance of
returns and [[beta].sub.2] is the coefficient of the asymmetric term,
capturing whether the impact of the shock is negative or positive.
[[beta].sub.3] represents the lagged volatility capturing persistency in
the conditional volatility of returns. To observe the impact of natural
disasters on the conditional volatility (ie, a measure of risk) of
insurance and composite stock markets, current and five lags of dummy
variables of natural disasters are also added in the conditional
variance equations.
DATA AND DESCRIPTIVE STATISTICS
We employ a financial data set and a natural disaster data set in
this study. All financial data series are obtained from the Global
financial data website. (5)
Data series employed in the insurance sector model include the
following:
(1) Japan TOPIX Insurance Index
(2) S&P 500 Insurance Composite
(3) 10-year bond yield in Japan
(4) 10-year bond yield in US
(5) US Dollar-Japanese Yen exchange rate
Data series employed in the composite stock market model contain
the following:
(1) Nikkei 225 Stock Average
(2) S&P 500 Composite Price Index
(3) 10-year bond yield in Japan
(4) 10-year bond yield in US
(5) US Dollar-Japanese Yen exchange rate
All data are daily over the period from 11 September 1989 to 8
April 2011. All stock returns are computed using the logarithm
difference of the share indices. Figure 1 presents the plots of stock
return series. Stock returns show similar patterns in the two countries.
However, Japanese stock returns are more volatile than stock returns in
the US. Also, Phillips-Perron tests unit root tests (not reported) were
conducted to make sure that all series are stationary.
The second data set is the natural disaster data set. The National
Geophysical Data Center provides data of several kinds of natural
disasters- earthquake, tsunami, and volcano eruptions around the world.
Information available is the time, location, and intensity. (6) The
JAXA/EORC Tropical Cyclone Database is used to manually identify
tropical cyclones that possibly impact Japan or the US, since tropical
cyclones usually last for several days and this database does not
provide specific date when a tropical cyclone hit a country. (7) We use
the date when a tropical cyclone is closest to a country and match it
with the stock market and other financial control variables. We examined
tropical cyclones in the North Western Pacific for Japan and those in
the North Eastern Pacific and North Atlantic for the US.
Four individual natural disaster (MD) dummy variables are employed
to represent each natural disaster event separately in estimations based
on the date it happened. Since it is likely that natural disasters may
happen on weekend or holidays, when stock markets do not open, the dates
of natural disasters are matched with the most recent trading day after
the disaster. For example, if an earthquake occurs on Saturday, the
performance of stock market next Monday is observed. There are totally
120 Japanese aggregate natural disaster events during 11 September 1989
to 8 April 2011, among which 61 are earthquakes, 36 are tsunamis, 4 are
volcano eruptions, and 55 are tropical cyclones. Among these natural
disasters, 36 earthquakes are associated with tsunamis. There are
totally 100 US aggregate natural disaster events during 11 September
1989 to 8 April 2011, among which 39 are earthquakes, 15 are tsunamis, 4
are volcano eruptions, and 51 are tropical cyclones. Among these natural
disasters, eight earthquakes are associated with tsunamis, and one
tsunami is associated with volcano eruption. There is no overlap between
tropical cyclones and other types of natural disasters in both Japan and
the US.
Descriptive statistics of insurance sector and composite stock
market returns are shown in Table 1. During the sample period, US stock
returns in both the insurance sector and the composite market are
positive and higher than those in Japan which are negative. Despite
differences in returns, the standard deviations of returns in both
countries are similar in the insurance sector, but the standard
deviation of the US composite stock returns is smaller than that of
Japan indicating that the Japanese insurance sector had a higher
volatility during the sample period. This is also reflected in the
differences in maximum and minimum returns in each sector. Skewness and
kurtosis figures indicate that stock returns are not normally
distributed and this is further supported by the Jarque-Bera tests,
which reject the normality hypothesis in all cases.
ESTIMATIONS AND FINDINGS
Before we report our empirical findings, we first discuss a problem
of time zone adjustment for the Japanese models. Hamao et al. (1990)
summarize trading hours of different stock markets when they investigate
the spillover effects across these markets. It reveals that there is no
overlapping trading time between the two markets. Since an international
date line divides the two stock markets so that Japan is one day ahead
of the US, the performance of the US stock market becomes overnight news
to the Japanese stock market. Therefore, there needs some data
pretreatment of the Japan TOPIX Insurance Index, S&P 500 Insurance
Composite, Nikkei 225 Stock Average and S&P 500 Composite Price
Index. Japan TOPIX Insurance Index at time t is matched with S&P 500
Insurance Composite at time (t-1). The Nikkei 225 Stock Average at time
t is matched with the S&P 500 index at time (t-1). Specially, Monday
in Japan is matched with Friday in the US. Although there were six
trading days per week in Japan before 1990, these observations are
deleted in order to be consistent with five trading days per week in the
US. Meanwhile, there is some inconsistency regarding holidays in the two
countries; these observations are also deleted when data are missing for
either stock market due to holidays. Finally, stock returns are
calculated after the data adjustment. In contrast, for the US model, the
performance of the Japanese stock market at time t is the overnight news
to the US stock market at time t; therefore, the S&P 500 Insurance
Composite and S&P index at time t are matched with Japan TOPIX
Insurance Index and the Nikkei 225 index at time t.
Table 2 presents the estimation results of both wealth and risk
effects of natural disasters. Panel A reports the wealth effects while
Panel B shows the risk effects. Panel C provides some diagnostic tests.
Wealth effects in the US
In Table 2, results for US composite market and insurance sector
returns show that movements in the Japanese stock market are significant
at the 1% level or better. A 10% increase in Japanese composite market
stock returns increases US composite market stock returns by 1.775%,
ceteris paribus, while a 10% increase in Japanese insurance sector stock
returns increases US insurance sector stock returns by 0.638%, ceteris
paribus. Domestic interest rate hikes have a negative impact on both US
composite market and insurance sector returns, while exchange rate
changes affect only US composite market returns. A 10% appreciation in
US dollar decreases US composite market return by 1.05% since
appreciation of US dollar discourages foreign capital flowing into US
stock market.
None of the natural disasters has any wealth effects on composite
stock market returns in the US, indicating that the market effectively
diversifies away the impact of natural disasters. For the US insurance
sector, among all natural disasters, only the volcanic eruption is
statistically significant at lag 3 and its impact is negative at the 1%
level or better. This negative coefficient shows that there is a 0.76%
daily drop in stock returns in the insurance sector three trading days
after the volcanic eruption is announced. The reason for late adjustment
might be that claims are made with some delay and/or it may take some
time to collect information about volcanic eruption's true damage.
Wealth effects in the Japan
Like the US market, none of the natural disasters have any wealth
effects on composite stock market returns in Japan, indicating that the
market diversifies away the impact of natural disasters. For the
Japanese insurance sector, all natural disasters, save the volcanic
eruption, have a significant impact. Lags 3 and 4 of both earthquakes
and tsunamis are statistically significant while tropical cyclone at the
event date is statistically significant. Three days following
earthquakes there is a 0.64% daily gain in stock returns in the
insurance sector but this is corrected the next day with a 0.56 % daily
loss and the net impact is positive. Three days following the tsunami
there is a 0.73 % daily gain in stock returns in the insurance sector
and gains continue the next day with an additional daily gain of 0.81%.
On the day of the tropical cyclones, there is a 0.34% daily gain in
stock returns in the insurance sector. The results for the Japanese
insurance sector are interesting because, besides significant adjustment
effects, they also indicate some supporting evidence for the
'gaining from loss hypothesis' associated with earthquakes,
tsunamis, and tropical cyclones.
Risk effects in the US
Panel B reports the results for the conditional variance or risk of
all returns for both the US and Japan. All beta coefficients are
statistically significant, indicating significant asymmetric effects of
shocks and quite persistency in the conditional variance. An important
question is whether natural disasters contribute to the movements in the
conditional variance of returns. For US composite market returns, only
statistically significant natural disaster variable that is of tropical
cyclones and at lag 1. After one day of tropical cyclones there is a
significant increase (0.71 units) in the conditional variance of US
composite market returns. In the US insurance sector, only natural
disaster variable that is significant is of volcanic eruptions and at
lag 3. After three days of volcanic eruptions there is a significant
increase (3.68 units) in the conditional variance of US insurance sector
returns.
Risk effects in Japan
For Japanese composite market returns, none of the natural disaster
variables has any statistically significant impact on the conditional
variance of the returns. In the Japanese insurance sector, both
earthquakes and tsunamis are associated with statistically significant
changes in the conditional variance of insurance sector returns. After
five days of earthquakes there is a significant increase (0.66 units) in
the conditional variance of Japanese insurance sector returns. After
four days of tsunamis there is a significant increase (1.22 units) in
the conditional variance of Japanese insurance sector returns. However,
in the next day there is a 0.79 units decline in the conditional
variance of Japanese insurance sector returns, indicating some market
adjustment.
Diagnostic tests reported in Panel C indicate all models do not
suffer from serial correlation and there are no remaining ARCH effects.
The only exception is the model for the Japanese insurance sector that
suffers from some remaining ARCH effects.
In summary, our results suggest that both US and Japanese composite
returns are not affected by any of natural disasters in our sample, but
the insurance sector in both countries are not able to diversify away
the impact of natural disasters. Regarding the risk effects, only the
Japanese composite market is immune to natural disasters, while the
conditional volatility of returns of Japanese insurance sector and
returns of both composite market and insurance sector in the US all are
affected by natural disasters.
Our results can be best compared with Worthington and Valadkhani
(2004) and Worthington (2008) for the Australian stock market.
Worthington and Valadkhani (2004) report significant wealth effects and
show that the net wealth effects can be positive and/or negative
depending on the adjustment in the days that follow natural disasters.
Our results for the Japanese insurance sector are quite consistent with
their findings.
Worthington (2008) reports no significant impact of natural
disasters on Australian stock market returns. The shocks do not create
systematic risk in the market, and they are diversified away at the
market level. Our results for Japanese and the US composite market
returns are consistent with this study.
CONCLUSIONS AND IMPLICATIONS
This study provides evidence on the impact of natural disasters on
both the insurance sector and the composite stock market in Japan and
the US. We find no wealth effects in both the US and Japanese composite
stock markets as returns in these markets are not affected by any of
natural disasters in our sample. However, there are significant wealth
effects in the insurance sector in both countries. The results for the
Japanese insurance sector indicate evidence for the 'gaining from
loss hypothesis' associated with earthquakes, tsunamis, and
tropical cyclones, while there is evidence against this hypothesis in
the US insurance sector. As far as the risk effects of natural
disasters, except the Japanese composite market, all returns are
affected by natural disasters.
There are also significant adjustment effects in returns in the
Japanese insurance sector. A possible reason for this finding is the
higher frequency of natural disasters in Japan compared with the US.
This study has important implications for investors since Japan and
the US face high probability of future natural disasters. Generally, our
results suggest that investors do not have to panic when there are
natural disasters in both Japan and the US, because the impact is
diversified away at the market level in both countries, although the
insurance sector in the US does suffer from natural disasters while the
Japanese insurance sector gains.
Acknowledgements
An earlier version of this paper was presented at the 2012
ACES-JACES-SSEM Pacific Rim Conference in Hawaii. Part of this paper was
completed when Kutan was on sabbatical and a visiting scholar at the
Center for Innovation and Competition-based Development Studies,
Bogazici University. Kutan also acknowledges the support by a TUBITAK
(The Scientific and Technological Research Council of Turkey) program.
The authors would like to thank Joe Brada (the Editor) for his very
useful comments and suggestions that greatly improved both the
exposition and quality of the paper. The usual disclaimer applies.
REFERENCES
Baklavaci, HF, Tunc, G and Aydogan, B. 2011: The impact of
firm-specific news on intraday dynamics: Evidence from the Turkish stock
market. Emerging Markets Finance and Trade 47(6): 99-119.
Chen, Z, Jiang, H and Donghui, L. 2010: Regulation change and
volatility spillovers: Evidence from China's stock markets.
Emerging Markets Finance and Trade 46(6): 140-157.
Enders, W. 2010: Applied econometric time series. Wiley: New York.
Hamao, Y, Masulis, RW and Ng, V. 1990: Correlations in price
changes and volatility across international stock markets. The Review of
Financial Studies 3(2): 281-307.
Janabi, MAMA, Hatemi, JA and Irandoust, M. 2010: Modeling
time-varying volatility and expected returns: Evidence from the GCC and
MENA regions. Emerging Markets Finance and Trade 46(5): 39-47.
Lamb, RP. 1995: An exposure-based analysis of property-liability
insurer stock values around Hurricane Andrew. The Journal of Risk and
Insurance 62(1): 111-123.
Ross, SA, Westerfield, RW and Jaffe, JF. 2010: Corporate finance.
McGraw- Hill/Irwin: New York.
Shelor, RM, Anderson, DC and Cross, ML. 1992: Gaining from loss:
Property- liability insurer stock values in the aftermath of the 1989
California earthquake. The Journal of Risk and Insurance 59(3): 476-488.
Worthington, A and Valadkhani, A. 2004: Measuring the impact of
natural disasters on capital markets: An empirical application using
intervention analysis. Applied Economics 36(19): 2177-2186.
Worthington, AC. 2008: The impact of natural events and disasters
on the Australian stock market: A GARCH-M analysis of storms, floods,
cyclones, earthquakes and bushfires. Global Business and Economics
Review 10(1): 1-10.
Yamori, N and Kobayashi, T. 2002: Do Japanese insurers benefit from
a catastrophic event? Market reactions to the 1995 Hanshin-Awaji
earthquake. Journal of the Japanese and International Economies 16(1):
92-108.
(1) Source: http://news.xinhuanet.com/english2010/world/2011-
04/25/c_13845359.html.
(2) Source: http://liberalsprinkles.blogspot.com/2011/03/facts-japan-earthquake- tsunami-nuclear.html.
(3) Source: http://www.bankrate.com/finance/insurance/top-10-costliest-natural- disasters-1.aspx.
(4) Baklavaci et al. (2011), Chen et al. (2010) and Janabi et al.
(2010) employ similar types of GARCH models.
(5) https://www.globalfinancialdata.com.
(6) http://www.ngdc.noaa.gov/hazard/hazards.shtml.
(7) http://sharaku.eorc.jaxa.jp/TYP_DB/index_e.shtml.
LIN WANG [1] & ALI M KUTAN [2]
[1] C.T. Bauer College of Business, University of Houston, Houston,
USA.
E-mail: lwang10@uh.edu
[2] Southern Illinois University Edwardsville, University Drive,
Edwardsville, IL 62025, USA, Borsa Istanbul, Istanbul, Turkey.
E-mail: akutan@siue.edu
Table 1: Descriptive statistics of stock returns
Insurance sector
stock return Composite stock return
Japan US Japan US
Mean -0.000197 0.000187 -0.000246 0.000260
Maximum 0.133957 0.215313 0.132346 0.124042
Minimum -0.166361 -0.142109 -0.121110 -0.094695
Std. Dev. 0.019246 0.019393 0.015947 0.012130
Skewness -0.068378 0.045142 -0.136468 -0.055774
Kurtosis 9.554346 13.67934 8.588657 12.75274
Jarque-Bera 9132.858 24237.00 6652.860 20214.80
Probability 0.000000 0.000000 0.000000 0.000000
Observations 5,100 5,100 5,100 5,100
Table 2: Natural disasters and stock returns
Variable US composite US insurance
Panel A--Conditional mean equation
Constant 0.0005 *** 0.0003 **
Foreign stock returns 0.1775 *** 0.0638 ***
Interest rate -0.0158 * -0.0248 *
Exchange rate -0.1050 *** -0.0162
Earthquake 0 0.0000 -0.0009
Lags 1 0.0000 0.0025
2 -0.0007 0.0010
3 0.0023 0.0001
4 0.0017 -0.0003
5 0.0010 -0.0026
Tsunami 0 -0.0022 -0.0008
Lags 1 -0.0020 -0.0053
2 0.0005 -0.0008
3 0.0005 -0.0038
4 -0.0011 0.0012
5 -0.0006 0.0000
Volcano eruption 0 -0.0034 0.0038
Lags 1 0.0034 -0.0002
2 0.0026 -0.0044
3 -0.0036 -0.0076 ***
4 -0.0075 0.0166
5 -0.0006 -0.0196
Tropical cyclone 0 -0.0026 -0.0006
Lags 1 -0.0019 0.0016
2 -0.0005 -0.0002
3 0.0009 -0.0004
4 -0.0023 0.0007
5 0.0000 -0.0003
Panel B--Conditional variance equation
[[beta].sub.0] -0.1855 *** -0.2076 ***
[[beta].sub.1] 0.1147 *** 0.1618 ***
[[beta].sub.2] -0.0666 *** -0.0552 ***
[[beta].sub.3] 0.9897 *** 0.9900 ***
Earthquake 0 -0.2836 0.0082
Lags 1 0.5896 0.1603
2 0.3219 -0.4301
3 -0.5869 0.2822
4 -0.3125 -0.0856
5 0.2101 0.0442
Tsunami 0 -0.7615 -0.1251
Lags 1 0.3544 -0.6263
2 -0.4665 0.2969
3 0.3249 -0.0198
4 -0.6269 0.0074
5 1.1523 0.4839
Volcano eruption 0 0.1773 0.7490
Lags 1 0.7234 -0.1913
2 -0.9673 0.0106
3 -2.1309 -2.9486
4 2.3135 3.6863 **
5 -0.2366 -0.6937
Tropical cyclone 0 -0.1635 0.1696
Lags 1 0.7149 ** -0.2583
2 -0.3591 0.4261
3 -0.3592 -0.4271
4 -0.0676 0.0221
5 -0.2554 0.0941
Panel C - Diagnostic tests
Log-likelihood 16813.3 14384.1
[Q.sup.1](10) Serial 10.35 8.02
Correlation
[Q.sup.2](10) Arch Effects 5.18 2.94
Japan Japan
Variable composite insurance
Panel A - Conditional mean equation
Constant -0.0003 *** -0.0006 ***
Foreign stock returns 0.5052 *** 0.0841 ***
Interest rate 0.0055 -0.0018
Exchange rate -0.0100 0.0175
Earthquake 0 0.0000 -0.0022
Lags 1 0.0003 -0.0001
2 0.0017 0.0021
3 0.0024 0.0064 **
4 0.0003 -0.0056 ***
5 0.0007 -0.0005
Tsunami 0 -0.0007 0.0027
Lags 1 -0.0006 -0.0013
2 -0.0008 -0.0019
3 -0.0015 0.0073 ***
4 0.0014 0.0081 **
5 0.0012 0.0028
Volcano eruption 0 -0.0133 0.0128
Lags 1 -0.0017 -0.0060
2 0.0014 -0.0093
3 -0.0020 -0.0043
4 -0.0042 0.0028
5 -0.0094 -0.0111
Tropical cyclone 0 0.0015 0.0034 *
Lags 1 0.0001 -0.0014
2 -0.0001 -0.0004
3 0.0007 0.0027
4 0.0015 0.0010
5 -0.0003 0.0003
Panel B - Conditional variance equation
[[beta].sub.0] -0.3859 *** -0.2880 ***
[[beta].sub.1] 0.1753 *** 0.1946 ***
[[beta].sub.2] -0.0879 *** -0.0511 **
[[beta].sub.3] 0.9709 *** 0.9827 ***
Earthquake 0 -0.0714 0.4999
Lags 1 0.3516 -0.3600
2 -0.2311 -0.2023
3 -0.1374 0.0063
4 -0.3016 -0.6986
5 0.2328 0.6690 **
Tsunami 0 -0.0190 -0.3920
Lags 1 0.0161 0.2671
2 -0.2242 -0.1137
3 0.1442 -0.2523
4 0.3216 1.2221 *
5 -0.3705 -0.7986 *
Volcano eruption 0 1.0014 0.0205
Lags 1 -1.5919 -0.2727
2 0.6601 0.5483
3 -2.2944 -0.0464
4 1.0534 -2.3648
5 0.9813 0.0537
Tropical cyclone 0 -0.1132 -0.2470
Lags 1 -0.1374 0.1791
2 0.2896 -0.0001
3 -0.0298 -0.0280
4 -0.3388 0.1243
5 0.2442 -0.1062
Panel C - Diagnostic tests
Log-likelihood 14896.1 13664.9
[Q.sup.1](10) Serial 4.82 7.13
Correlation
[Q.sup.2](10) Arch Effects 8.62 21.77 ***
Notes:
***, **, * denote significance level at 1%, 5%, and 10% levels,
respectively.
Foreign stock returns, the interest rate, and the exchange rate are
expressed in terms of one-day lag to deal with the potential
endogeneity bias.
Foreign stock returns refer to the US and Japan composite
(insurance) index returns when the dependent variables are Japan
and US composite (insurance) index returns, respectively.
All equations include up to five lags of the dependent variable to
deal with the autocorrelation in the residuals. These results are
not reported due to space considerations.
