Description
The objective of this research is to assess whether two events, the 9-21 Earthquake in
1999 and the Severe Acute Respiratory Syndrome outbreak in 2003, had a temporary or long-term
impact on the inbound tourism demand from Japan. Furthermore, a comparative study is conducted to
assess whether intervention analysis produces better forecasts compared with forecasts without
intervention analysis.
International Journal of Culture, Tourism and Hospitality Research
Forecasting Japanese tourism demand in Taiwan using an intervention analysis
J ennifer C.H. Min
Article information:
To cite this document:
J ennifer C.H. Min, (2008),"Forecasting J apanese tourism demand in Taiwan using an intervention analysis",
International J ournal of Culture, Tourism and Hospitality Research, Vol. 2 Iss 3 pp. 197 - 216
Permanent link to this document:http://dx.doi.org/10.1108/17506180810891582
Downloaded on: 24 January 2016, At: 22:05 (PT)
References: this document contains references to 56 other documents.
To copy this document: [email protected]
The fulltext of this document has been downloaded 1431 times since 2008*
Users who downloaded this article also downloaded:
Austin Uzama, (2009),"Marketing J apan's travel and tourism industry to international tourists",
International J ournal of Contemporary Hospitality Management, Vol. 21 Iss 3 pp. 356-365 http://
dx.doi.org/10.1108/09596110910948341
Oscar Claveria, J ordi Datzira, (2010),"Forecasting tourism demand using consumer expectations", Tourism
Review, Vol. 65 Iss 1 pp. 18-36http://dx.doi.org/10.1108/16605371011040889
Lan-Lan Chang, Kenneth F. Backman, Yu Chih Huang, (2014),"Creative tourism: a preliminary examination
of creative tourists’ motivation, experience, perceived value and revisit intention", International J ournal
of Culture, Tourism and Hospitality Research, Vol. 8 Iss 4 pp. 401-419http://dx.doi.org/10.1108/
IJ CTHR-04-2014-0032
Access to this document was granted through an Emerald subscription provided by emerald-srm:115632 []
For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald for
Authors service information about how to choose which publication to write for and submission guidelines
are available for all. Please visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and practice to the benefit of society. The company
manages a portfolio of more than 290 journals and over 2,350 books and book series volumes, as well as
providing an extensive range of online products and additional customer resources and services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the Committee
on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archive
preservation.
*Related content and download information correct at time of download.
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Forecasting Japanese tourism
demand in Taiwan using
an intervention analysis
Jennifer C.H. Min
Tourism Department, Hsing Wu College, Taipei, Taiwan
Abstract
Purpose – The objective of this research is to assess whether two events, the 9-21 Earthquake in
1999 and the Severe Acute Respiratory Syndrome outbreak in 2003, had a temporary or long-term
impact on the inbound tourism demand from Japan. Furthermore, a comparative study is conducted to
assess whether intervention analysis produces better forecasts compared with forecasts without
intervention analysis.
Design/methodology/approach – The data adopted in this study consist of monthly visitor
arrivals from Japan to Taiwan for the period January 1979-September 2006. The ?rst 321 observations
( January 1979-September 2005) are used to develop two tentative models, with and without
intervention analyses, and then compare with the known values (October 2005-September 2006) for
accuracy testing.
Findings – Experimental results show that the effect of both disasters on Japanese inbound tourism
presented only temporarily, and the forecasting ef?ciency of ARIMA with intervention is superior to
that of a model without intervention.
Research limitations/implications – The study had dif?culty accurately delineating the rebound
in Japanese tourist based on monthly data. There are other factors that might in?uence a rebound,
such as people’ fading memories or the purpose of visitation. The geographic proximity of Taiwan to
Japan could also account for perceived risk factors.
Practical implications – The results indicate that the Japanese inbound arrivals sharply dropped
following both of the two disastrous occurrences, suggesting that the Japanese tourists are likely to be
responsive to prompt marketing strategies and messages. The practical implication for tourism
operators include the usefulness of reinforcing the package holiday by establishing an attractively
priced travel package or offering a package with a variety of highly desirable or unique features to
increase competition.
Originality/value – This study is a ?rst attempt in the tourism literature to model Japanese demand
for travel to Taiwan after these two traumatic crises.
Keywords Tourism, Disasters, Forecasting, Taiwan, Japan
Paper type Research paper
1. Introduction
Since the 1980s, Japan has witnessed an enormous growth in outbound tourism due to
its economic strength and the signi?cant appreciation of the yen (Sakai et al., 2000;
Polunin, 1989; Murakami and Go, 1990; Cha et al., 1995; Ahmed and Krohn, 1992;
Mak et al., 2005; Lim and McAleer, 2005). This emergence of Japan as a leading tourist
generating country has been a highlight of world tourism in recent years. In addition,
because of the high-spending tendency of Japanese tourists (Choy, 1998; Nozawa, 1992;
Murphy and Williams, 1999), ranking as the top tourism spender in the Asia Paci?c
and fourth among all countries in international tourism expenditure in 2003-2005,
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/1750-6182.htm
Forecasting
Japanese tourism
in Taiwan
197
Received November 2006
Revised March 2007
Accepted December 2007
International Journal of Culture,
Tourism and Hospitality Research
Vol. 2 No. 3, 2008
pp. 197-216
qEmerald Group Publishing Limited
1750-6182
DOI 10.1108/17506180810891582
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
after Germany, the USA and the UK (World Tourism Organization, 2006), Japan has
been a signi?cant market sector for many tourist destinations. Recognizing the
importance of Japanese tourists to international tourism in terms of numbers and
expenditure, considerable research has been undertaken regarding the structure of
Japanese outbound tourism industries and the characteristics of Japanese travelers
(Tokuhisa, 1980; Sage, 1985; Ahmed and Krohn, 1992; Cha et al., 1995; March, 1997a;
Holtzman et al., 1991; Iverson, 1997b; Lang et al., 1993; Balaz and Mitsultake, 1998;
Pinhey and Iverson, 1994; Heung et al., 2001). Moreover, several studies have attempted
to compare Japanese tourist behavior across national boundaries (Dybka, 1988; March,
1997b; Iverson, 1997a; Chen, 2000; Mihalik et al., 1993; Wang and Lim, 2005; Lim and
Wang, 2005). The purpose of these tasks has been to identify the pro?table segment
that might be tapped through more effective approaches.
The relationship between Japan and Taiwan is an especially close one. Taiwan was
a colony of Japan from 1895 to 1945 prior to the KMT (Kuo-Ming Ton) Party’s ?ight to
Taiwan to exercise its sovereignty. During that time, only the Japanese language was
allowed to be spoken and learned by residents; therefore, the elder generations of
Taiwanese can speak ?uent Japanese. Besides, with many common features in the two
nations’ cultures and history, Taiwan has long been a favorite destination for Japanese
tourists (Lin, 1990). As one of the major tourist markets for Taiwan, Japan has
accounted for over 30 percent of international tourist arrivals to the island since the
very early stage of Taiwan’s international tourism development. Figure 1 shows
Taiwan’s market share of international tourist arrivals from Japan since 1979.
According to Figure 1, the market share of tourist arrivals from Japan maintained
a steady share in the 1980s.The Japanese population in this period enjoyed an improved
lifestyle with increasing leisure and disposable income. This pattern had a dramatic effect
on demand for overseas travel and Taiwan greatly bene?ted from this trend. However,
Taiwan’s market share fromJapan decreased slightly after 1993, because of higher living
costs inTaiwan, and a severe economic recession inJapanfromthe early1990s. The Asian
?nancial crisis that began in mid-1997 also had far-reaching consequences for the
economy of Japan, including its outbound tourism industry. The decline in the Japanese
market also coincided with a decrease in Taiwan’s overall tourism market share during
1997-2000. Inparticular, the Severe Acute RespiratorySyndrome (SARS) outbreak in2003
resulted in the lowest recorded point for the inbound market from Japan, at 14.8 percent.
Figure 1.
Taiwan’s market share
of international tourist
arrivals from Japan,
1979/1-2006/9
0.0%
10.0%
20.0%
30.0%
40.0%
50.0%
60.0%
70.0%
80.0%
1
9
7
9
1
9
8
1
1
9
8
3
1
9
8
5
1
9
8
7
1
9
8
9
1
9
9
1
1
9
9
3
1
9
9
5
1
9
9
7
1
9
9
9
2
0
0
1
2
0
0
3
2
0
0
5
Year
M
a
r
k
e
t
s
h
a
r
e
IJCTHR
2,3
198
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
This is because most Japanese inbound visitors have pleasure as the main reason for their
trip rather than business, so it is easy to cancel a trip in the face of environmental changes.
In terms of international tourism market share, Japan has long been Taiwan’s major
market and despite the above, the overall numbers of Japanese tourists to the island have
not slowed down. Japanese visitors for the year of 2005 even reached a new record of
1.1 million, double-digit growth compared to 2004.
Because of the perishable nature of the products and services tourism supplies,
forecasting plays a signi?cant role for destination countries’ tourism related decision
making. It is not possible to store most tourism products, such as unoccupied hotel
rooms, unsold airline seats, unused concert halls, or vacant dining rooms, etc. Tourism
forecasts therefore are essential to tourism planning and resource allocation.
Forecasting tourist numbers is helpful in reducing losses caused by disparities
between the demand and supply of tourism goods and services (Law and Au, 1999).
Moreover, reliable and accurate forecasting is needed to assist tourism decision makers
plan more effectively and ef?ciently. At the same time, tourism demand is highly
sensitive to catastrophic and negative in?uences, such as natural disasters and other
extreme events. Tourists undoubtedly suffer the most serious impact from negative
events because they are unfamiliar with local remedial resources that can be relied on
to avoid serious consequences (Faulkner, 2001).
Although Japan has been Taiwan’s largest tourist generating country for a long time,
no research has been undertaken of forecast demand from Japan. More critically, no
tourismliterature has examined the impact of two events, the September 21st Earthquake
(also known as the 9-21 Earthquake or just 921) in 1999 or the SARS outbreak in 2003,
on Japanese inbound tourism. These two calamities were the largest natural disaster in
the 20th and the ?rst catastrophic epidemic in the twenty-?rst century in Taiwan,
respectively. Although these two events have quite different features as calamities (one is
a natural disaster, the other an epidemic), both of themcaused a big decline in the number
of tourist arrivals. Thus, the objective of this research is to assess whether the unexpected
tragedies of the earthquake and SARS had a temporary or long-term impact on Japanese
inbound tourism demand. To account for these two events in a Japanese inbound model,
the empirical analysis has two stages. Firstly, an intervention analysis is performed to
provide an estimated impact of these disasters on the inbound tourism from Japan.
Secondly, a comparative study is conducted to assess whether intervention analysis
produces better forecasts compared with forecasts without intervention analysis.
The paper is organized in the following manner: Section 1 presents the motives and
objectives of the study. Section 2 illustrates the detrimental consequences of the
9-21 Earthquake and the SARS outbreak, and provides a detailed description of how
international tourism and Japanese inbound markets have been affected, and how the
government and tourism industry of Taiwan responded to such threats in order to
regain Japanese tourist numbers. Data and measurement issues are outlined in Section
3. Section 4 is devoted to data analysis, and the presentation of both the seasonal
ARIMA(SARIMA) and SARIMAwith intervention models. The ?nal section, Section 5,
summarizes the study and presents conclusions.
2. Background
Two events in recent times greatly affected the tourism industry in Taiwan. The ?rst
was a strong earthquake on September 21, 1999, the second the SARS outbreak in 2003.
Forecasting
Japanese tourism
in Taiwan
199
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Needless to say, during and after these events the island’s economy was greatly
impacted, with the worst damage to inbound tourism where tourist numbers were
drastically reduced.
2.1 The 9-21 Earthquake, 1999
This earthquake, measuring 7.3 on the Richter scale, killed over 2,400 people, injured
more than 8,000, and left in excess of 100,000 homeless. The greatest effects of the
9-21 Earthquake were felt instantly by the tourism industry, since it damaged tourist
facilities at numerous popular destinations. Much infrastructure serving the island
residents and tourists was damaged or destroyed. The worst damage was in central
Taiwan, near the epicenter in Nantou County. Nearby theme parks, such as the
Formosan Aboriginal Culture Village, Sun Moon Lake, and Li Shan (Pear Mountain),
were hurt badly during the peak season for tours in the quarter. In addition to scenic
areas, a large number of listed ancient monuments and unlisted historic buildings were
also harshly damaged. Some historical relics away from the faults also suffered
seriously.
According to the Tourism Bureau report, the number of visitors to 230 major scenic
spots dropped by 27 percent for the September-December period of 1999. The room
occupancy rates of hotels for international tourist plummeted by an average of about
60 percent, and international airline reservation cancellations soared to 210,000 at
the same period. Statistics showed that government-operated scenic spots suffered
losses amounting to approximately US$19.5 million, and privately operated tourist
enterprises suffered losses of about US$119 million (Tourism Bureau, 2000).
Huang and Min (2002) examined the recovery of visitor arrivals to Taiwan after the
9-21 Earthquake. The results indicate that Taiwan had not fully recovered from the
sharply reduced inbound arrivals nearly a year after the disastrous earthquake, even
though there was large expenditure on infrastructure rehabilitation. For a long period,
the island had not experienced such an earthquake as the one on September 21.
Since Japan has long been the leading generator of international tourism for
Taiwan, no effort can be spared toward regaining the Japanese tourist market. The
placement of “Peace of Mind” advertisements in major Japanese media was increased,
and promotion seminars were held jointly with large Japanese air carriers in Tokyo,
Fukuoka, Osaka, and Nagoya. Related posters were placed throughout Tokyo railways
and bus stations. Furthermore, a marketing program of “Go, Taiwan” was carried out
in cooperation with China Airlines and Japan Asia Airways. Nevertheless, to attract
tourists to Taiwan, there was still a need to substantially reduce the overall prices of
taking a trip to Taiwan during the early recovery stage. There were signi?cant fare
discounts for Japan-Taiwan routes. Most of the hotels in Taiwan also made many
efforts in response to declining visitor numbers. Many of Taiwan’s international tourist
hotels mailed promotional materials to every previous Japanese guest to try to rescue a
sharply declining Japanese tourist market (Travel Trend News, 1999). In addition to
carrying out measures to revitalize the Japanese tourist market in the aftermath of the
quake, the Taiwan Tourism Bureau also placed 11 advertisements in major Japanese
media during the year 2000 to draw Japanese tourists to visit Taiwan. Moreover, each
Japanese arrival was presented with a box of Alpine Oolong tea while traveling in the
central part of the island to increase the visitation.
IJCTHR
2,3
200
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
2.2 The SARS outbreak, 2003
From the end of 2002 onwards, SARS spread rapidly through the medium of
international travel to more than 26 countries on ?ve continents. Within months of its
?rst appearance in mainland China, SARS had resulted in 8,422 reported probable
cases, of which 916 resulted in deaths. Taiwan had to cope with the third largest SARS
outbreak, with a total of 674 contracted cases and 84 deaths. The largest outbreaks
occurred in Mainland China (5,327 cases and 349 deaths) and Hong Kong (1,755 cases
and 300 deaths) (World Health Organization – WHO, 2003a).
SARS is an acute respiratory illness caused by infection by the corona virus spread
predominantly by droplets and direct or indirect contact. On March 12, 2003, due to the
spread of the illness to several countries in a short period of time, WHO issued a global
alert, suggesting that national authorities implement heightened surveillance for cases
of SARS. Three days later, WHO named the mysterious illness after its symptoms:
SARS, and issued a series of emergency guidelines for travelers and airlines
(WHO, 2003b). The purpose of the announcement was to eliminate the spread of
infection by international travel.
The ?rst case of SARS in Taiwan was a 54-year-old businessman, who had traveled
in Guangdong Province, China, where the ?rst case of SARS is known to have occurred
(WHO, 2003c). The subsequent spread of the epidemic led to WHO designating Taiwan
an affected area on March 15, 2003. Because of the lack of knowledge about the disease,
people had no good information to help them deal with the virus, and tourism
organizations also lacked the resources to protect themselves at home or means of
getting help from abroad. In May, the situation got even worse with the SARS virus
spreading at an alarming rate and the number of new cases and deaths increasing. Fear
and uncertainty blanketed the island. Taipei, the capital of Taiwan, was placed on the
WHO travel advisory list on May 8, and travel warnings were extended to the entire
island on May 21. This was the strictest world-wide travel advisory issued by WHO in
its 55-year history. Following these announcements, 30 nations imposed additional
restrictions on travelers from Taiwan, and up to 48 nations also listed Taiwan as a
travel destination to be avoided (Tourism Bureau, 2003). Consequently, the number of
tourist arrivals fell signi?cantly, reaching levels the island had never witnessed before.
During this time, the Taiwan Government established methods and procedures for
SARS education, prevention and treatment. The effect of the administration’s
decisions, though gradual, began to be seen, and the number of infected cases and
deaths gradually fell. As a result, WHO lifted Taiwan from its SARS travel advisory
list on June 17, 27 days after putting Taiwan on the list, and on July 5 removed it from
the list of areas with recent local transmission.
The SARS epidemic undermined Taiwan’s tourism industry and impeded the ?ow
of tourists. According to a Tourism Bureau report, international arrival numbers fell
sharply in the second quarter by 71.54 percent from a year earlier due to the SARS
outbreak. These declines slowed to 19.73 and 11.1 percent in the third and fourth
quarter, respectively, (Tourism Bureau, 2004). Since Japan is Taiwan’s main source of
visitors, the government and private sector jointly rolled out various tourism
promotion measures and new tour destinations to attract Japanese tourists to the island
after July 5, when WHO removed Taiwan from its list of SARS-affected regions.
Advertising was carried out for tourism promotions through major print and electronic
media channels in Japan. Post-SARS promotions emphasized the safety of travel in the
Forecasting
Japanese tourism
in Taiwan
201
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
island under the themes “Exhausted in Japan, Rest Up in Taiwan,” and “Take it Easy.”
The Taiwan Visitors Association was commissioned to work with domestic tour
operators to form tourism promotional events for the Japanese market. In addition,
media representatives, travel writers and tour operators were invited to take part in
tours from July to September to highlight the safety of traveling in Taiwan after SARS.
The Tourism Bureau sent a delegation to hold promotional events in Tokyo, Osaka
and Nagoya in August to inform the Japanese public about the post-SARS recovery
status of Taiwan. Moreover, several special tour packages were introduced, such as
10,000 yen travel coupon books awarded by draw to 5,000 Japanese travelers. These
swift countermeasures were adopted to revitalize the Japanese market and encourage
Japanese tourism back to the island.
3. Method
3.1 Data collection
Tourist arrivals are de?ned in this study as persons traveling to Taiwan from Japan,
which is not usual residence, for at least 24 hours up to a period not exceeding
12 months. In this research, the Japanese inbound tourism demand series adopted are
unadjusted monthly tourist arrivals from Japan to Taiwan for the period from January
1979 to September 2006, and this set of data has been obtained from the Monthly
Report on Tourism published by the Tourism Bureau of Taiwan. The statistical
package for SAS/ETS 9.1 version software is employed to develop and use the
SARIMA with and without intervention models.
3.2 Time series modeling
Time series analysis can be a valuable tool for forecasters to view trends in data,
both long-term and cyclical. The Box-Jenkins model is the most common time series
technique to be applied. In 1970, Box and Jenkins made the ARIMA model popular and
easy to use by proposing a model building methodology including four stages:
identi?cation, estimation, diagnostic checking and forecasting. This methodology also
allows researchers not only to discover hidden patterns in data but also to generate
forecasts. The model has been providing benchmark forecasts for more than three
decades, and has been fruitfully applied in economics, business, ?nance, engineering,
and many other branches of the physical and social sciences.
The univariate ARIMA model with intervention (also known as intervention
analysis) of a dependent variable is subjected to the effects of exceptional events on a
time series model. It was presented by Box and Tiao (1975) to examine the reduction of
oxidant pollution levels in downtown Los Angeles by adopting two interventions: the
diversion of traf?c by the opening of the Golden State Freeway and the coming into
effect of a new law. Since Box and Tiao ?rst investigated the level shift or drift in the
time series within the framework of the Box-Jenkins transfer function modeling,
intervention analysis has been employed widely to estimate various types of economic
or environmental impacts through policy changes, strikes, promotions, disasters and
the like (Chang and Lin, 1997; Goh and Law, 2002; Kapombe and Colyer, 1999; Lai and
Lu, 2005; Sharma and Khare, 1999; Goh, 2005; Min et al., 2006).
Intervention analysis has been well-documented in modeling and forecasting
tourism demand. Bonham and Gangnes (1996) adopted time series intervention
analysis to evaluate ex post the effect of the 1987 Hawaii hotel room tax imposition.
IJCTHR
2,3
202
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Empirical ?ndings show that there was no statistically signi?cant evidence of any
impact by the tax. Lee et al. (2005) used the intervention model to estimate the impact of
the September 11 terrorist attacks on US air transport passenger demand and to see
how long it took the US air transport industry to recover from the crisis. Coshall (2003)
applied the intervention model to assessing the impact of three interventions, the US
bombing of Libya in 1986, the Lockerbie air disaster in 1988, and the Persian Gulf crisis
during 1990-1991, on the ?ow of UK air passengers to a variety of destinations. In
another study (Coshall, 2005), used the intervention approach to measure the impact of
different events, namely terrorism, war, and the foot and mouth disease outbreak, on
expenditure by visitors to the UK and UK tourists abroad. Min et al. (2006) used a
SARIMA with intervention model both to evaluate the impact of the SARS outbreak in
Taiwan and incorporated the intervention into the outbound time series model to
improve parameter estimates and forecasts. Studies by Goh and Law (2002) and Lai
and Lu (2005), compared the SARIMA with intervention model with different
techniques; the results showed that it outperformed all other models when signi?cant
intervention in the series existed. Based on this literature review, the following
hypothesis is tested in this study:
H1. Intervention analysis of Japanese tourism demand in Taiwan after the
9-21 Earthquake and the SARS outbreak produces better forecasts in
comparison with forecasts without intervention analysis.
3.3 Mathematical formulation
In univariate ARIMA analysis, the time series to be modeled is based on observations
generated by an underlying process to be identi?ed. In addition, one of the basic
principles of model building is that the model should adequately represent the data
using as few parameters as possible (Box and Jenkins, 1970). This section is a brief
review formulation of the time series model.
The goal is to ?nd a good model that adequately represents the process-generating
mechanism. There are two useful representations to express the behavior of observed
time series processes, namely the autoregressive (AR) and moving average (MA)
models. The AR model is used to describe a time series in which the current
observation depends on its past values, whereas the MA model is used to describe a
time series process as a linear function of current and past random shocks. The sample
partial autocorrelation function and the sample autocorrelation function (ACF) are
adopted to identify the models in AR and MA, respectively. The general formulation
suggested by Box and Jenkins as an autoregressive moving average can be written as:
fðBÞZ
t
¼ uðBÞa
t
ð1Þ
where Z
t
and a
t
represent the Japanese inbound numbers and random error terms at
time t, respectively. B is a backward shift operator de?ned by BZ
t
¼ Z
t21
. f(B) and
u(B) are AR and MA operators of orders p and q, respectively, and are expressed as:
fðBÞ ¼ 1 2f
1
B 2f
1
B 2f
3
B 2. . . 2f
p
B
p
ð2Þ
and:
uðBÞ ¼ 1 2u
1
B 2u
2
B 2u
3
B 2· · · 2u
q
B
q
ð3Þ
Forecasting
Japanese tourism
in Taiwan
203
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
If a time series presents nonstationary, it can often be transformed into a stationary
series by taking ?rst difference. Nonstationary models after differencing can be
expressed as:
fðBÞð1 2BÞ
d
Z
t
¼ uðBÞa
t
ð4Þ
(1 2 B)
d
Z
t
means d of difference of the process to be stationary, which also can be
written as ARIMA( p, d, q).
Tourist arrivals series often exhibits a seasonal periodic component which
measures at regular calendar intervals within a year. Box and Jenkins have generalized
the ARIMA model to deal with seasonality and de?ne it as SARIMA. It is denoted as
ARIMA( p, d, q )x(P,D,Q)
s
in which p and P are the orders of the AR operator; d and D
are the differencing; and q and Qare the orders of the MA of non-seasonal and seasonal
components, respectively. It combines the ARIMA( p, d, q ) and yields:
fðBÞFðB
s
Þð1 2BÞ
d
ð1 2B
s
Þ
D
Z
t
¼ uðBÞQðB
s
Þa
t
ð5Þ
For intervention analysis, it is to study the impact of exceptional external events as:
y
t
¼
X
k
i¼1
v
i
ðBÞ
d
i
ðBÞ
B
b
i
j
ti
þ N
t
ð6Þ
where v
i
ðBÞ ¼ v
i0
2v
i1
B 2· · · 2v
is
i
B
s
i
and d
i
ðBÞ ¼ 1 2d
i1
B 2· · · 2d
ir
i
B
r
i
are the
numerator and denominator polynomial of the intervention effects function,
respectively. j
ti
is represented by a series of 0’s and 1’s denoting non-occurrence and
occurrence of the intervention. B is the backshift operator, de?ned as
By
t
¼ y
t21
and B
b
y
t
¼ y
t2b
. N
t
is the noise series which represents the background
observed series y
t
without the intervention effects.
3.4 Types of intervention responses
Box and Tiao (1975) considers the intervention variables as having a pulse or step
function. If the effects of an intervention are temporary or transient and will die out
after time t, a pulse function is used, where:
j
t
¼ P
ðTÞ
t
¼
0
1
(
t – T
t ¼ T
ð7Þ
If the effects of an intervention are expected to remain permanently after time t to some
extent, a step function is adopted, where:
j
t
¼ S
ðTÞ
t
¼
0
1
(
t , T
t $ T
ð8Þ
In addition, Box and Tiao illustrated some typical transfer functions for step and pulse
response patterns shown in Figure 2. The following Figure 2(a) can be used as a step or
abrupt permanent effect with level of magnitudev. Figure 2(b) shows the gradual
permanent effect with change rate d and long-run response level v/1 2 d. Figure 2(c) is
a special case of Figure 2(b) as d ¼ 1. Figure 2(d) shows an abrupt and temporary effect
IJCTHR
2,3
204
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
with a gradual decay of rate d that will approach the pre-intervention level with no
permanent effect. Figure 2(e) is similar to Figure 2(d), but with a permanent effect of
magnitude v
2
. Figure 2(e) is a linear combination of Figure 2(a) with Figure 2(d), the
pattern of Figure 2(a) is equivalent to ðvB=1 2BÞP
ðTÞ
t
through the relationship
ð1 2BÞS
ðTÞ
t
¼ P
ðTÞ
t
. Figure 2(f) can be used instead of Figure 2(e) if an initial impact is
considered.
4. Empirical results
Intervention analysis begins by identifying a plausible set of ARIMA models.
Generally, the longest data span of the pre- and post-intervention observations is used
to identify N
t
(Ender, 2004; Coshall, 2003). Therefore, the data adopted in this study
consist of monthly visitor arrivals from Japan during successive months over a 27-year
period ( January 1979-September 2006). Among the 333 observations, the ?rst
321 observations ( January 1979-September 2005) are used to develop two tentative
models, with and without intervention analyses. These two models are employed to
Figure 2.
Forms of intervention
responses
w B S
t
(T)
STEP:
w
i
(B)
d
i
(B)
S
t
(T)
PULSE:
w
i
(B)
d
i
(B)
P
t
(T)
w B
1 – d (B)
S
t
(T)
w B
1 – d (B)
P
t
(T)
w B
1 – B
S
t
(T)
w
w
1
w
1
w
0
w
1
w
2
w
2
w
1–d
w
(b)
(c)
(a) (d)
(e)
(f)
w
1
B
1 – d B
P
t
(T)
w
2
B
1 – B
+
w
1
B
w
0
1 – d B
P
t
(T)
w
2
B
1 – B
+
Forecasting
Japanese tourism
in Taiwan
205
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
forecast the 12 future values of Japanese inbound arrivals of October 2005-September
2006, and then compare with the known values for accuracy testing.
The ?rst aspect of a time series analysis is to plot the data; Figure 3 shows this
point. The plot exhibits some interesting features. First, an attention is immediately
drawn to an obvious decline from the 292nd observation (April, 2003), which is caused
by the SARS outbreak. Another clear reduction can also be noticeably identi?ed from
the 248th observation (September, 1999), relates to the 9-21 Earthquake. Second, an
initial plot of the data of Japanese visitor arrivals reveals irregular variation and
seasonal variation. The two disasters in this study, the 9-21 Earthquake and the SARS
epidemic, have a profound in?uence on the Japanese tourism demand for travel to
Taiwan, and both are intervention events in the Taiwan’s tourism.
4.1 Pre-intervention model
The appropriate model for noise component N
t
is identi?ed prior to the known
interventions, the 9-21 Earthquake and SARS, and the ?rst 248 observations, January
1979 through August 1999, are utilized for the identi?cation process. There is a
question of whether the series is stationary or not. The time series plot of the ?rst
248 observations is clearly nonstationary. To con?rm this, the ACF and augmented
Dickey-Fuller test (ADF) are used to test for stationarity in the available time-series
data. For the ADF test, the null hypothesis is that the series are nonstationary and
therefore include a unit root. The representations and hypothesis of the ADF test are as
follows:
Dy
t
¼ b
0
þ dt þ b
1
y
t21
þ
X
p
i¼1
g
i
Dy
t2i
þ 1
t
ð9Þ
An ADF regression investigates the null hypothesis of non-stationarity (b
1
¼ 0)
against the alternative of stationarity (b
1
, 0).
The time series represents nonstationary by the ADF test. ACF also shows positive
and high for a number of lags; and decays fairly slowly. It indicates that the ?rst
248 time series is nonstationary. Thus, differencing is necessary. However, the use of
(1 2 B) alone does not remove the effects of nonstationarity from the data, and the ACF
Figure 3.
Time series plot on
Japanese visitor arrivals
from January 1979 to
September 2006
0
20,000
40,000
60,000
80,000
1,00,000
1,20,000
1
9
7
9
1
9
8
1
1
9
8
3
1
9
8
5
1
9
8
7
1
9
8
9
1
9
9
1
1
9
9
3
1
9
9
5
1
9
9
7
1
9
9
9
2
0
0
1
2
0
0
3
2
0
0
5
Year
N
u
m
b
e
r
o
f
J
a
p
a
n
e
s
e
a
r
r
i
v
a
l
s
Intervention Events
IJCTHR
2,3
206
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
at lags 12, 24, 36, and 48 shows the most distinct seasonal effect, indicating the
presence of seaonality in the given time series. Hence, seasonal differencing is
warranted. In view of the above indications, the ?rst and seasonal difference
(1 2 B)(1 2 B
12
), should be taken in order to achieve stationariy. The pattern is
indicative of a multiplicative MA(1) and MA(12) model, that is (1 2 u
1
B)(1 2 u
12
B
12
)
after the ACF diagrams and ADF tests, and the Ljnug-Box Q statistic for residual
checking. Following the identi?cation, the estimate for the Japanese tourist of the
pre-intervention is yielded by Maximum Likelihood Estimation as:
ð1 2BÞð1 2B
12
ÞZ
t
¼ ð1 20:54451BÞð1 20:60876B
12
Þa
t
ð10Þ
4.2 SARIMA model with intervention
The Japanese inbound series in this study indicates that the ?rst impact of the
9-21 Earthquake and SARS interventions starts to take effect in September 1999 and
April 2003, respectively. The intervention variables j
t1
and j
t2
are set to value 1 for
September 1999 and April 2003 and 0 otherwise. According to Figure 3, the effect of the
single impulse lasts more than one month due to the SARS impact, a second drop in
May 2003 suggests a numerator lag and the following exponential increase suggests
a denominator lag in the transform function (Brocklebank and Dickey, 2003). As the
impact of the earthquake is illustrated as an abrupt, a temporary impact (shown
in Figure 3), we expect the 9-21 Earthquake has a similar type of intervention
as the SARS incident. Therefore, the two intervention models are set in the form
ðv
o1
2v
11
B=1 2d
11
BÞj
t1
and ðv
o2
2v
12
B=1 2d
12
BÞj
t2
.
The Maximum Likelihood estimates of the SARIMA model with intervention are:
Z
t
¼
v
o1
2v
11
B
1 2d
11
B
j
t1
þ
v
o2
2v
12
B
1 2d
12
B
j
t2
þ
ð1 2u
1
BÞð1 2u
12
B
12
Þ
ð1 2BÞð1 2B
12
Þ
a
t
Z
t
¼
20:20654 20:44164B
1 20:80722B
j
t1
þ
20:64039 21:95801B
1 20:63207B
j
t2
þ
ð1 20:60420BÞð1 20:65753B
12
Þ
ð1 2BÞð1 2B
12
Þ
a
t
ð11Þ
Since Dubin –Watson test is designed to detect ?rst order AR errors, it is therefore
used to examine if autocorrelation is present. It can be expressed as:
d ¼
X
n
t¼2
^ 1
t
2 ^ 1
t21
ð Þ
2
=
X
n
t¼1
^ 1
2
t
ð12Þ
where:
^ 1
t
¼ Y
t
2
^
b
0
2
^
b
1
X
1t
2
^
b
2
X
2t
ð13Þ
If the actual errors e
t
are uncorrelated, the numerator of d has an expected value of
approximately2(n 2 1)s
2
andthe denominator has anexpectedvalue of about ns
2
. Thus,
with errors uncorrelated, the ratio d should be about 2 (Brocklebank and Dickey, 2003).
The result of Dubin-Watson test in the intervention model obtains 2.0054, indicating the
errors are uncorrelated.
Forecasting
Japanese tourism
in Taiwan
207
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Table I presents the results of the estimation and shows that all estimate parameters
are signi?cant with very small p-values. The plot of the estimated residual ACFis shown
in Figure 4, and it is evident fromthe ?gure that the residuals are not autocorrelated, and
reveal approximate white noise. Moreover, the Ljung-Box Q statistic is not signi?cant
and indicates the SARIMA model with intervention does ?t well.
Parameter Estimate Standard error t-value Approx Pr . jtj Lag Variable
u
1
0.60420 0.04496 13.44 ,0.0001 1 MA(1)
u
12
0.65753 0.04490 14.64 ,0.0001 12 MA(12)
v
o1
20.20654 0.08581 22.41 0.0161 0 9-21Earthquake
v
11
0.44164 0.09048 4.88 ,0.0001 1 9-21Earthquake
d
11
0.80722 0.05920 13.64 ,0.0001 1 9-21Earthquake
v
o2
20.64039 0.13916 24.60 ,0.0001 0 SARS
v
12
1.95801 0.09058 21.62 ,0.0001 1 SARS
d
12
0.63207 0.02963 21.33 ,0.0001 1 SARS
Table I.
Full-intervention model
by maximum likelihood
estimation
Figure 4.
Estimated residual ACF
on ARIMA model with
intervention
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
–1 –0.5 0 0.5 1
Correlation coefficients
Autocorrelations
IJCTHR
2,3
208
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
The permanent effect of the earthquake impact can be investigated by adding one level
shift variable L
t
, and de?ning the level shift variable value to be 1 prior to April 2003
and 0 otherwise. The coef?cient of permanent effect l
2
¼ 0.26047, with standard error
0.13537, is not signi?cant. This means the permanent effect is not signi?cantly
different from 0, indicating no permanent effect after the SARS event. This is
equivalent to the term {v
1
B=1 2B}P
ðTÞ
t
, with permanent effect v
1
, and the transfer
function suggested by Box and Tiao (1975) as following:
v
0
þ
v
1
B
1 2dB
þ
v
2
B
1 2B
P
ðTÞ
t
ð14Þ
The in?uence of change rate d in event of SARS, impacting the effect of transfer
function I
t
of the Japanese inbound series, is considered in the intervention model
as follows:
I
t
¼
v
o2
2v
12
B
1 2d
12
B
j
t
¼ ðv
o2
2v
12
BÞð1 þd
12
B þd
2
12
B
2
þd
3
12
B
3
þ . . .Þj
t
¼ v
o2
j
t
þ ðv
o2
d
12
2v
12
Þj
t21
þd
12
ðv
o2
d
12
2v
12
Þj
t22
þ d
2
12
ðv
o2
d
12
2v
12
Þj
t23
þ . . .
ð15Þ
The same procedure to test the permanent effect of the earthquake is also applied.
The results show that the coef?cient of permanent effect l
1
¼ 20.08413, with
standard error 0.22510, is not signi?cant either, indicating no permanent effect after the
earthquake.
4.3 ARIMA model
As a ?rst step to model identi?cation, the plot of the data for the monthly Japanese
tourist arrivals from January 1979 to September 2005 is considered. The series is
checked for stationarity using ACF and ADF for unit roots and, if necessary, the series
is transformed by taking appropriate differences to render the series stationary. The
t-statistics of ADF test on the series at level are larger than the critical value at the
5 percent signi?cance level. Thus, the null hypothesis of unit root is not rejected,
indicating the series is nonstationary. Differencing is necessary. Since the series exhibits
varying seasonal patterns, it is essential to test for the presence of seasonal unit roots.
After seasonal difference (1 2 B
12
), the series achieves stationarity. Various SARIMA
models are ?tted to the Japanese arrivals series after seasonal differencing. Since the
Akaike information criterion (AIC) and the Schwarz Bayesian criterion (SBC) are
generally used in model selection criteria, whereby the smaller values are preferred;
thus, the selected models are then compared according to the AIC and SBC. These
procedures are repeated until anappropriate model is found. After best model is selected,
the Dubin-Watson test is adopted to obtain 1.9994, demonstrating the errors are
uncorrelated. Furthermore, residual checking using p-value associated with the
Ljnug-Box Qstatistic shows that the residual of the model represents random, indicating
the residuals approximate white noise. The ACF of the residual series is also used to
check residuals (Figure 5). The ACF is clean and reveals potential existent white noise
and no anomalies. Following the identi?cation, the estimate for the SARIMA without
intervention model is obtained by Maximum Likelihood Estimation as:
Forecasting
Japanese tourism
in Taiwan
209
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
ð1 20:67361BÞð1 2B
12
ÞZ
t
¼ ð1 þ 0:36322BÞð1 20:73315B
12
Þa
t
ð16Þ
4.4 Evaluation of two models
The two models are estimated using the data from January 1979 to September 2005.
Accuracy measurement of the two approaches for the period October 2005 to
September 2006 is based on mean absolute deviation (MAD), mean absolute percentage
error (MAPE) and mean square error (MSE), which are often adopted for the
performance of each model. Each of these measurements is de?ned as follows:
MAD ¼
1
n
X
n
i¼1
Diff
i
j j ð17Þ
MAPE ¼
1
n
X
n
i¼1
Diff
i
j j
Act
i
£ 100 percent ð18Þ
Figure 5.
Estimated residual
ACF on ARIMA model
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
–1 –0.5 0 0.5 1
Correlation coefficients
Autocorrelations IJCTHR
2,3
210
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
MSE ¼
1
n
X
n
i¼1
Diff
2
i
ð19Þ
Table II shows the summary of SARIMA and intervention models evaluation, and
Figure 6 shows a graphical presentation of the values generated by the two forecasting
methods.
As Table II shows, the forecasting output from the intervention analysis is more
accurate, with a relatively small amount of error. The minimum values of MAD, MAPE
SARIMA SARIMA with intervention
Variable (1,0,1)(0,1,1) (0,1,1)(0,1,1)
MA(1) u
1
20.363
* *
0.571
* *
MA(12) u
12
0.733
* *
0.660
* *
AR(1) f
1
0.674
* *
9-21earthquake v
o1
20.209
*
9-21earthquake v
11
0.435
* *
9-21earthquake 0.812
* *
SARS v
o2
20.865
* *
SARS v
12
1.983
* *
SARS d
12
0.663
* *
AIC 2273.3 2532.6
SBC 2262.1 2502.8
Residuals Q(36) 0.888 0.220
R
2
0.807 0.785
MAD(2005/10-2006/09) 0.2302 0.0561
MAPE(2005/10-2006/09) 2.01 percent 0.49 percent
MSE(2005/10-2006/09) 0.0671 0.0052
Notes: Signi?cant levels at:
*
0.05 and
* *
0.01, respectively
Table II.
Evaluation of SARIMA
and intervention models
Figure 6.
Graphical presentation
of two forecasting
methods
Actural visitors (1,0,1)(0,1,1) (0,1,1)(0,1,1)with intervention
50,000
60,000
70,000
80,000
90,000
1,00,000
1,10,000
1,20,000
Forecasting month
N
u
m
b
e
r
o
f
J
a
p
a
n
e
s
e
v
i
s
i
t
o
r
s
t
o
T
a
i
w
a
n
1 2 3 4 5 6 7 8 9 10 11 12
Forecasting
Japanese tourism
in Taiwan
211
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
and MSE show that the deviations between the predicted values derived by the
intervention analysis and the actual values are smaller. The R
2
of the intervention
model is also larger. In addition, based on the criterions of AIC and SBC,
the intervention model is better than the model of SARIMA without intervention.
As Figure 6 shows, the forecast data from the intervention model are closer to the
actual data. In view of the above-empirical examinations, intervention model
outperforms the SARIMA without intervention in describing the Japanese tourist ?ows
after the events of the 9-21 Earthquake and SARS. The statistical results not only
support the hypothesis, but are also consistent with previous studies on intervention
analysis in the tourism literature.
5. Implications and conclusions
This study adopts an intervention model to estimate the impacts of two disasters, the
9-21 Earthquake and SARS, and to explore if they had a temporary or long-term impact
on Japanese inbound tourism demand. The intervention analysis is compared to a
model without intervention to see which technique can accurately predict Japanese
tourist ?ows to Taiwan under the in?uence of a drastic environmental change such as
an earthquake or epidemic. The forecasting performance is evaluated using
out-of-sample predictions. According to the ?ndings, the effect of both disasters on
Japanese inbound tourism presents only temporarily. Experimental results also
indicate that the forecasting ef?ciency of ARIMA with intervention is superior to that
of a model without intervention.
The results indicate that the Japanese inbound arrivals sharply dropped following
both of the two disastrous occurrences, suggesting that the Japanese tourists are likely
to be responsive to prompt marketing strategies and messages. Since Japanese tourists
are more adverse to risky leisure pursuits (Polunin, 1989; Ahmed and Krohn, 1992), the
safety of accommodations and recreational facilities, and the security of the
environment should be strongly advertised to help the Japanese minimize the feeling of
uncertainty. By doing this, the Japanese tourists can be overcome without diluting
interest in traveling to Taiwan. Moreover, the propensity of package tour travel is
strong among Japanese tourists (Lim and McAleer, 2005). The practical implication for
tourism operators include the usefulness of reinforcing the package holiday by
establishing an attractively priced travel package or offering a package with a variety
of highly desirable or unique features to increase competition. Furthermore, because
Japanese tourists are more at ease in groups (Ahmed and Krohn, 1992; Pinhey and
Iverson, 1994), tour packages should promote group activities rather than organizing
activities for them in isolation from other people, in order to ful?ll their need for
security. Printed advertisements and packages offered through travel agents could
emphasize the fact that tourism destinations in Taiwan are still functional in order to
rekindle Japanese visitors’ interest.
This study is a ?rst attempt in the tourism literature to model Japanese demand for
travel to Taiwan after these two traumatic crises. However, there are some limitations
that could be overcome in future research. This study had dif?culty accurately
delineating the rebound in Japanese tourist based on monthly data. It may be more
accurate to classify recovery periods using weekly data in future studies. In addition,
tourismresearchers can apply other forecasting techniques to produce optimal results in
studies of the effects of unexpected misfortunes in the future. Since arti?cial neural
IJCTHR
2,3
212
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
networks are able to outperformothers in most situations (Law, 2001; Burger et al., 2001),
these can be used to investigate which generate more accurate forecasts. Also, a possible
future study would be to test the accuracy of a vector autoregressive model capable of
producing accurate medium to long-term forecasts, which does not require the
generation of forecasts for explanatory variables before forecasts of dependent variables
can be obtained (Song and Witt, 2006). Another suggested direction for further research
is the ARMAX modeling approach, which is a single equation econometric technique
that accounts for the dynamic relationship between variables such as GDP and
exchange rates. Accurate predictions from such forecasting techniques can aid tourist
authorities and professionals in Taiwan improve their planning and decision-making
processes. Lastly, there are other factors that might in?uence a rebound, such as people’
fading memories or the purpose of visitation. The geographic proximity of Taiwan to
Japan could also account for perceived risk factors. Whether these aspects represent
unexamined factors in this study could become the basis of further studies.
The value of forecasting lies in its ability to assist tourism authorities make
operational, tactical and strategic decisions (Wang and Lim, 2005; Law and Au, 1999).
Therefore, in deciding which forecasting method to use after a negative occurrence,
practitioners and policy makers may con?dently apply intervention analysis.
Although both these events impeded the ?ow of Japanese tourists, traf?c almost
returned to pre-catastrophe levels afterwards. Thus, examining such effects in a
detailed and accurate manner enables tourism authorities to learn and appreciate the
dynamics of tourist behavior for future negative occurrences to achieve more
sophisticated forward planning. This research will potentially bene?t researchers of
tourism and tourism authorities in Taiwan as well as worldwide.
References
Ahmed, Z.U. and Krohn, F.B. (1992), “Understanding the unique consumer behavior of Japanese
tourists”, J. of Travel & Tourism Market, Vol. 3 No. 1, pp. 73-86.
Balaz, V. and Mitsultake, M. (1998), “Japanese tourists in transition countries of Central Europe:
present behavior and future trends”, Tourism Manage, Vol. 19 No. 5, pp. 433-43.
Bonham, C.S. and Gangnes, B. (1996), “Intervention analysis with cointegrated time series:
the case of the Hawaii hotel room tax”, Appl. Econ., Vol. 28 No. 10, pp. 1281-93.
Box, G.E.P. and Jenkins, G.M. (1970), Time Series Analysis: Forecasting and Control, Holden-Day
Inc., San Francisco, CA.
Box, G.E.P. and Tiao, G.C. (1975), “Intervention analysis with applications to economic and
environmental problems”, J. of the Am. Stat. Associ., Vol. 70, pp. 70-4.
Brocklebank, J. and Dickey, D.A. (2003), SAS for Forecasting Time Series, SAS Institute Inc.,
Cary, NC.
Burger, C.J.S.C., Dohnal, M., Kathrada, M. and Law, R. (2001), “A practitioners guide to
time-series methods for tourism demand forecasting – a case study of Durban,
South Africa”, Tourism Manage, Vol. 22 No. 4, pp. 403-9.
Cha, S., McCleary, K.W. and Uysal, M. (1995), “Travel motivations of Japanese overseas travelers:
a factor-cluster segmentation approach”, J. of Travel. Res., Vol. 34 No. 1, pp. 33-9.
Chang, N. and Lin, Y.T. (1997), “An analysis of recycling impacts on solid waste generation by
time series intervention modeling”, Resour, Cons. and Recycling, Vol. 19 No. 3, pp. 165-85.
Forecasting
Japanese tourism
in Taiwan
213
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Chen, J.S. (2000), “Cross-cultural differences in travel information acquisition among tourists
from three Paci?c-rim countries”, J. of Hospitality &TourismRes., Vol. 24 No. 2, pp. 239-51.
Choy, D.J.L. (1998), “Changing trends in Asia-Paci?c tourism”, Tourism Manage, Vol. 19 No. 4,
pp. 381-2.
Coshall, J. (2003), “The threat of terrorism as an intervention on international travel ?ows”, J. of
Travel. Res., Vol. 42 No. 1, pp. 4-12.
Coshall, J. (2005), “Interventions on UK earnings and expenditures overseas”, Ann. of Tourism
Res., Vol. 32 No. 5, pp. 592-609.
Dybka, J. (1988), “Overseas travel to Canada: new research on the perceptions and preferences of
the pleasure travel market”, J. of Travel Res., Vol. 26 No. 4, pp. 12-15.
Ender, W. (2004), Applied Econometrics Time Series, Willey, New York, NY.
Faulkner, B. (2001), “Towards a framework for tourism disaster management”, Tourism
Manage, Vol. 22 No. 2, pp. 135-47.
Goh, B.H. (2005), “The dynamic effects of the Asian ?nancial crisis on construction demand and
tender price levels in Singapore”, Building and Environ, Vol. 40 No. 2, pp. 267-76.
Goh, C. and Law, R. (2002), “Modeling and forecasting tourism demand for arrivals with
stochastic nonstationary seasonality and intervention”, Tourism Manage, Vol. 23 No. 5,
pp. 499-510.
Heung, V.C.S., Qu, H. and Chu, R. (2001), “The relationship between vacation factors and
socio-demographic and traveling characteristics: the case of Japanese leisure travelers”,
Tourism Manage, Vol. 22 No. 3, pp. 259-69.
Holtzman, W.L., Murthy, B. and Gordon, J.C. (1991), “Cultural bridging with the Japanese”,
Cornel Hotel and Rest Admin Quart, Vol. 32 No. 3, pp. 52-9.
Huang, J.H. and Min, J.C.H. (2002), “Earthquake devastation and recovery in tourism: the Taiwan
case”, Tourism Manage, Vol. 23 No. 2, pp. 145-54.
Iverson, T.J. (1997a), “Decision timing: a comparison of Korean and Japanese travelers”, Int J. of
Hospitality Manage, Vol. 16 No. 2, pp. 209-19.
Iverson, T.J. (1997b), “Japanese visitors to Guam: lessons from experience”, J. of Travel &
Tourism Market, Vol. 6 No. 1, pp. 41-54.
Kapombe, C.M. and Colyer, D.A. (1999), “Structural time series analysis of US broiler exports”,
Agri. Econ., Vol. 21 No. 3, pp. 295-307.
Lai, S.L. and Lu, W.L. (2005), “Impact analysis of September 11 on air travel demand in the USA”,
J. of Air Trans. Manage, Vol. 11 No. 6, pp. 455-8.
Lang, C.T., O’Leary, J.T. and Morrison, A.M. (1993), “Activity segmentation of Japanese female
overseas travelers”, J. of Travel & Tourism Market, Vol. 2 No. 4, pp. 1-22.
Law, R. (2001), “The impact of the Asian ?nancial crisis on Japanese demand for travel to
Hong Kong: a study of various forecasting techniques”, J. of Travel & Tourism Market,
Vol. 10 Nos 2/3, pp. 47-65.
Law, R. and Au, N. (1999), “A neural network model to forecast Japanese demand for travel to
Hong Kong”, Tourism Manage, Vol. 20 No. 1, pp. 89-97.
Lee, S., Oh, C. and O’Leary, J.T. (2005), “Estimating the impact of the September 11 terrorist
attacks on the US air transport passenger demand using intervention analysis”, Tourism
Anal., Vol. 9 No. 4, pp. 355-61.
Lim, C. and McAleer, M. (2005), “Analyzing the behavioral trends in tourist arrivals from Japan
to Australia”, J. of Travel. Res., Vol. 43 No. 4, pp. 414-21.
IJCTHR
2,3
214
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Lim, C. and Wang, Y. (2005), “Forecasting Japanese outbound tourism using seasonal time series
models”, Int. J. of Hospitlaiy Manage., Vol. 3 No. 2, pp. 37-54.
Lin, M. (1990), “Taiwan: magnet for Japanese tourists”, Cornel Hotel and Rest Admin Quart,
Vol. 31 No. 1, pp. 96-7.
Mak, J., Carlile, L. and Dai, S. (2005), “Impact of population aging on Japanese international travel
to 2025”, J. of Travel. Res., Vol. 44 No. 2, pp. 151-76.
March, R. (1997a), “An exploratory study of buyer-supplier relationships in international
tourism: the case of Japanese wholesalers and Australian suppliers”, J. of Travel &
Tourism Market, Vol. 6 No. 1, pp. 55-68.
March, R. (1997b), “Diversity in Asian outbound travel industries: a comparison between
Indonesia, Thailand, Taiwan, South Korea and Japan”, Int. J. of Hospitality Manage.,
Vol. 16 No. 2, pp. 231-8.
Mihalik, B.J., Uysal, M. and Pan, M.C. (1993), “A comparison of information sources used by
vacationing Germans and Japanese”, Hospitality Res. J., Vol. 18 No. 3, pp. 39-46.
Min, J., Wu, B. and Wu, K. (2006), “Forecasting outbound tourism with intervention analysis”,
The Bus Review, Vol. 6 No. 2, pp. 179-86.
Murakami, K. and Go, F. (1990), “Transitional corporations capture Japanese market”, Tourism
Manage, Vol. 11 No. 4, pp. 348-52.
Murphy, A. and Williams, P.W. (1999), “Attracting Japanese tourists into the rural hinterland:
implications for rural development and planning”, Tourism Manage, Vol. 20 No. 4,
pp. 487-99.
Nozawa, H.A. (1992), “Marketing analysis of Japanese outbound travel”, Tourism Manage,
Vol. 13 No. 2, pp. 226-33.
Pinhey, T.K. and Iverson, T.J. (1994), “Safety concerns of Japanese visitors to Guam”, J. of Travel
& Tourism Market, Vol. 3 No. 2, pp. 87-94.
Polunin, I. (1989), “Japanese travel boom”, Tourism Manage, Vol. 10 No. 1, pp. 4-8.
Sage, C. (1985), “Concerning the Japanese market”, Tourism Manage, Vol. 7 No. 4, pp. 301-3.
Sakai, M., Brown, J. and Mak, J. (2000), “Population aging and Japanese international travel in the
21st century”, J. of Travel Res., Vol. 38 No. 3, pp. 212-20.
Sharma, P. and Khare, M. (1999), “Application of intervention analysis for assessing the
effectiveness of CO pollution control legislation in India”, Trans. Res. Part D, Vol. 4 No. 6,
pp. 427-32.
Song, H. and Witt, S.F. (2006), “Forecasting international tourist ?ows to Macau”, Tourism
Manage, Vol. 7 No. 2, pp. 214-24.
Tokuhisa, T. (1980), “Tourism within, from and to Japan”, Int. Social Sci. J., Vol. 32 No. 1,
pp. 128-50.
Tourism Bureau (2000), 1999 Annual Report on Tourism, Tourism Bureau, Taipei.
Tourism Bureau (2003), “The status of tourism industry in?uenced by SARS”, available at:http://anonymouse.ws/cgi-bin/anon-www.cgi/http:/202.39.225.133/sars/c04.htm
Tourism Bureau (2004), 2003 Annual Report on Tourism, Tourism Bureau, Taipei.
Travel Trend News (1999), “1999 special report: marketing strategies for inbound and outbound
markets after the 9-21 Earthquake”, Travel Trend News, Vol. 209, pp. 16-27.
Wang, Y. and Lim, C. (2005), “Assessing forecasting models for outbound Japanese travellers to
Australia and China”, J. of Appl. Econ. and Manage., Vol. 3 No. 2, pp. 1-15.
Forecasting
Japanese tourism
in Taiwan
215
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
WHO (2003a), “Summary table of SARS cases by country”, World Health Organization,
November 1, 2002-August 7, 2003, available at: www.who.int/csr/sars/country/2003_08_
15/en/f (accessed August 15, 2006).
WHO (2003b), “World Health Organization issues emergency travel advisory”, World Health
Organization, available at: www.who.int/csr/sars/archive/2003_03_15/en/ (accessed
March 15, 2006).
WHO (2003c), “Taiwan, China: transmission interrupted in last outbreak area”, World Health
Organization, available at: www.who.int/csr/don/2003_07_05/en/ (accessed July 5, 2006).
World Tourism Organization (2006), “UNWTO world tourism barometer”, available at: www.
unwto.org/facts/menu.html (accessed September 6, 2006).
Corresponding author
Jennifer C.H. Min can be contacted at: [email protected]
IJCTHR
2,3
216
To purchase reprints of this article please e-mail: [email protected]
Or visit our web site for further details: www.emeraldinsight.com/reprints
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
This article has been cited by:
1. Tianshu Zheng, John Farrish, Miranda Kitterlin. 2016. Performance Trends of Hotels and Casino Hotels
Through the Recession: An ARIMA With Intervention Analysis of Stock Indices. Journal of Hospitality
Marketing & Management 25, 49-68. [CrossRef]
2. Bo Peng, Haiyan Song, Geoffrey I. Crouch. 2014. A meta-analysis of international tourism demand
forecasting and implications for practice. Tourism Management 45, 181-193. [CrossRef]
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
doc_325107960.pdf
The objective of this research is to assess whether two events, the 9-21 Earthquake in
1999 and the Severe Acute Respiratory Syndrome outbreak in 2003, had a temporary or long-term
impact on the inbound tourism demand from Japan. Furthermore, a comparative study is conducted to
assess whether intervention analysis produces better forecasts compared with forecasts without
intervention analysis.
International Journal of Culture, Tourism and Hospitality Research
Forecasting Japanese tourism demand in Taiwan using an intervention analysis
J ennifer C.H. Min
Article information:
To cite this document:
J ennifer C.H. Min, (2008),"Forecasting J apanese tourism demand in Taiwan using an intervention analysis",
International J ournal of Culture, Tourism and Hospitality Research, Vol. 2 Iss 3 pp. 197 - 216
Permanent link to this document:http://dx.doi.org/10.1108/17506180810891582
Downloaded on: 24 January 2016, At: 22:05 (PT)
References: this document contains references to 56 other documents.
To copy this document: [email protected]
The fulltext of this document has been downloaded 1431 times since 2008*
Users who downloaded this article also downloaded:
Austin Uzama, (2009),"Marketing J apan's travel and tourism industry to international tourists",
International J ournal of Contemporary Hospitality Management, Vol. 21 Iss 3 pp. 356-365 http://
dx.doi.org/10.1108/09596110910948341
Oscar Claveria, J ordi Datzira, (2010),"Forecasting tourism demand using consumer expectations", Tourism
Review, Vol. 65 Iss 1 pp. 18-36http://dx.doi.org/10.1108/16605371011040889
Lan-Lan Chang, Kenneth F. Backman, Yu Chih Huang, (2014),"Creative tourism: a preliminary examination
of creative tourists’ motivation, experience, perceived value and revisit intention", International J ournal
of Culture, Tourism and Hospitality Research, Vol. 8 Iss 4 pp. 401-419http://dx.doi.org/10.1108/
IJ CTHR-04-2014-0032
Access to this document was granted through an Emerald subscription provided by emerald-srm:115632 []
For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald for
Authors service information about how to choose which publication to write for and submission guidelines
are available for all. Please visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and practice to the benefit of society. The company
manages a portfolio of more than 290 journals and over 2,350 books and book series volumes, as well as
providing an extensive range of online products and additional customer resources and services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the Committee
on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archive
preservation.
*Related content and download information correct at time of download.
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Forecasting Japanese tourism
demand in Taiwan using
an intervention analysis
Jennifer C.H. Min
Tourism Department, Hsing Wu College, Taipei, Taiwan
Abstract
Purpose – The objective of this research is to assess whether two events, the 9-21 Earthquake in
1999 and the Severe Acute Respiratory Syndrome outbreak in 2003, had a temporary or long-term
impact on the inbound tourism demand from Japan. Furthermore, a comparative study is conducted to
assess whether intervention analysis produces better forecasts compared with forecasts without
intervention analysis.
Design/methodology/approach – The data adopted in this study consist of monthly visitor
arrivals from Japan to Taiwan for the period January 1979-September 2006. The ?rst 321 observations
( January 1979-September 2005) are used to develop two tentative models, with and without
intervention analyses, and then compare with the known values (October 2005-September 2006) for
accuracy testing.
Findings – Experimental results show that the effect of both disasters on Japanese inbound tourism
presented only temporarily, and the forecasting ef?ciency of ARIMA with intervention is superior to
that of a model without intervention.
Research limitations/implications – The study had dif?culty accurately delineating the rebound
in Japanese tourist based on monthly data. There are other factors that might in?uence a rebound,
such as people’ fading memories or the purpose of visitation. The geographic proximity of Taiwan to
Japan could also account for perceived risk factors.
Practical implications – The results indicate that the Japanese inbound arrivals sharply dropped
following both of the two disastrous occurrences, suggesting that the Japanese tourists are likely to be
responsive to prompt marketing strategies and messages. The practical implication for tourism
operators include the usefulness of reinforcing the package holiday by establishing an attractively
priced travel package or offering a package with a variety of highly desirable or unique features to
increase competition.
Originality/value – This study is a ?rst attempt in the tourism literature to model Japanese demand
for travel to Taiwan after these two traumatic crises.
Keywords Tourism, Disasters, Forecasting, Taiwan, Japan
Paper type Research paper
1. Introduction
Since the 1980s, Japan has witnessed an enormous growth in outbound tourism due to
its economic strength and the signi?cant appreciation of the yen (Sakai et al., 2000;
Polunin, 1989; Murakami and Go, 1990; Cha et al., 1995; Ahmed and Krohn, 1992;
Mak et al., 2005; Lim and McAleer, 2005). This emergence of Japan as a leading tourist
generating country has been a highlight of world tourism in recent years. In addition,
because of the high-spending tendency of Japanese tourists (Choy, 1998; Nozawa, 1992;
Murphy and Williams, 1999), ranking as the top tourism spender in the Asia Paci?c
and fourth among all countries in international tourism expenditure in 2003-2005,
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/1750-6182.htm
Forecasting
Japanese tourism
in Taiwan
197
Received November 2006
Revised March 2007
Accepted December 2007
International Journal of Culture,
Tourism and Hospitality Research
Vol. 2 No. 3, 2008
pp. 197-216
qEmerald Group Publishing Limited
1750-6182
DOI 10.1108/17506180810891582
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
after Germany, the USA and the UK (World Tourism Organization, 2006), Japan has
been a signi?cant market sector for many tourist destinations. Recognizing the
importance of Japanese tourists to international tourism in terms of numbers and
expenditure, considerable research has been undertaken regarding the structure of
Japanese outbound tourism industries and the characteristics of Japanese travelers
(Tokuhisa, 1980; Sage, 1985; Ahmed and Krohn, 1992; Cha et al., 1995; March, 1997a;
Holtzman et al., 1991; Iverson, 1997b; Lang et al., 1993; Balaz and Mitsultake, 1998;
Pinhey and Iverson, 1994; Heung et al., 2001). Moreover, several studies have attempted
to compare Japanese tourist behavior across national boundaries (Dybka, 1988; March,
1997b; Iverson, 1997a; Chen, 2000; Mihalik et al., 1993; Wang and Lim, 2005; Lim and
Wang, 2005). The purpose of these tasks has been to identify the pro?table segment
that might be tapped through more effective approaches.
The relationship between Japan and Taiwan is an especially close one. Taiwan was
a colony of Japan from 1895 to 1945 prior to the KMT (Kuo-Ming Ton) Party’s ?ight to
Taiwan to exercise its sovereignty. During that time, only the Japanese language was
allowed to be spoken and learned by residents; therefore, the elder generations of
Taiwanese can speak ?uent Japanese. Besides, with many common features in the two
nations’ cultures and history, Taiwan has long been a favorite destination for Japanese
tourists (Lin, 1990). As one of the major tourist markets for Taiwan, Japan has
accounted for over 30 percent of international tourist arrivals to the island since the
very early stage of Taiwan’s international tourism development. Figure 1 shows
Taiwan’s market share of international tourist arrivals from Japan since 1979.
According to Figure 1, the market share of tourist arrivals from Japan maintained
a steady share in the 1980s.The Japanese population in this period enjoyed an improved
lifestyle with increasing leisure and disposable income. This pattern had a dramatic effect
on demand for overseas travel and Taiwan greatly bene?ted from this trend. However,
Taiwan’s market share fromJapan decreased slightly after 1993, because of higher living
costs inTaiwan, and a severe economic recession inJapanfromthe early1990s. The Asian
?nancial crisis that began in mid-1997 also had far-reaching consequences for the
economy of Japan, including its outbound tourism industry. The decline in the Japanese
market also coincided with a decrease in Taiwan’s overall tourism market share during
1997-2000. Inparticular, the Severe Acute RespiratorySyndrome (SARS) outbreak in2003
resulted in the lowest recorded point for the inbound market from Japan, at 14.8 percent.
Figure 1.
Taiwan’s market share
of international tourist
arrivals from Japan,
1979/1-2006/9
0.0%
10.0%
20.0%
30.0%
40.0%
50.0%
60.0%
70.0%
80.0%
1
9
7
9
1
9
8
1
1
9
8
3
1
9
8
5
1
9
8
7
1
9
8
9
1
9
9
1
1
9
9
3
1
9
9
5
1
9
9
7
1
9
9
9
2
0
0
1
2
0
0
3
2
0
0
5
Year
M
a
r
k
e
t
s
h
a
r
e
IJCTHR
2,3
198
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
This is because most Japanese inbound visitors have pleasure as the main reason for their
trip rather than business, so it is easy to cancel a trip in the face of environmental changes.
In terms of international tourism market share, Japan has long been Taiwan’s major
market and despite the above, the overall numbers of Japanese tourists to the island have
not slowed down. Japanese visitors for the year of 2005 even reached a new record of
1.1 million, double-digit growth compared to 2004.
Because of the perishable nature of the products and services tourism supplies,
forecasting plays a signi?cant role for destination countries’ tourism related decision
making. It is not possible to store most tourism products, such as unoccupied hotel
rooms, unsold airline seats, unused concert halls, or vacant dining rooms, etc. Tourism
forecasts therefore are essential to tourism planning and resource allocation.
Forecasting tourist numbers is helpful in reducing losses caused by disparities
between the demand and supply of tourism goods and services (Law and Au, 1999).
Moreover, reliable and accurate forecasting is needed to assist tourism decision makers
plan more effectively and ef?ciently. At the same time, tourism demand is highly
sensitive to catastrophic and negative in?uences, such as natural disasters and other
extreme events. Tourists undoubtedly suffer the most serious impact from negative
events because they are unfamiliar with local remedial resources that can be relied on
to avoid serious consequences (Faulkner, 2001).
Although Japan has been Taiwan’s largest tourist generating country for a long time,
no research has been undertaken of forecast demand from Japan. More critically, no
tourismliterature has examined the impact of two events, the September 21st Earthquake
(also known as the 9-21 Earthquake or just 921) in 1999 or the SARS outbreak in 2003,
on Japanese inbound tourism. These two calamities were the largest natural disaster in
the 20th and the ?rst catastrophic epidemic in the twenty-?rst century in Taiwan,
respectively. Although these two events have quite different features as calamities (one is
a natural disaster, the other an epidemic), both of themcaused a big decline in the number
of tourist arrivals. Thus, the objective of this research is to assess whether the unexpected
tragedies of the earthquake and SARS had a temporary or long-term impact on Japanese
inbound tourism demand. To account for these two events in a Japanese inbound model,
the empirical analysis has two stages. Firstly, an intervention analysis is performed to
provide an estimated impact of these disasters on the inbound tourism from Japan.
Secondly, a comparative study is conducted to assess whether intervention analysis
produces better forecasts compared with forecasts without intervention analysis.
The paper is organized in the following manner: Section 1 presents the motives and
objectives of the study. Section 2 illustrates the detrimental consequences of the
9-21 Earthquake and the SARS outbreak, and provides a detailed description of how
international tourism and Japanese inbound markets have been affected, and how the
government and tourism industry of Taiwan responded to such threats in order to
regain Japanese tourist numbers. Data and measurement issues are outlined in Section
3. Section 4 is devoted to data analysis, and the presentation of both the seasonal
ARIMA(SARIMA) and SARIMAwith intervention models. The ?nal section, Section 5,
summarizes the study and presents conclusions.
2. Background
Two events in recent times greatly affected the tourism industry in Taiwan. The ?rst
was a strong earthquake on September 21, 1999, the second the SARS outbreak in 2003.
Forecasting
Japanese tourism
in Taiwan
199
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Needless to say, during and after these events the island’s economy was greatly
impacted, with the worst damage to inbound tourism where tourist numbers were
drastically reduced.
2.1 The 9-21 Earthquake, 1999
This earthquake, measuring 7.3 on the Richter scale, killed over 2,400 people, injured
more than 8,000, and left in excess of 100,000 homeless. The greatest effects of the
9-21 Earthquake were felt instantly by the tourism industry, since it damaged tourist
facilities at numerous popular destinations. Much infrastructure serving the island
residents and tourists was damaged or destroyed. The worst damage was in central
Taiwan, near the epicenter in Nantou County. Nearby theme parks, such as the
Formosan Aboriginal Culture Village, Sun Moon Lake, and Li Shan (Pear Mountain),
were hurt badly during the peak season for tours in the quarter. In addition to scenic
areas, a large number of listed ancient monuments and unlisted historic buildings were
also harshly damaged. Some historical relics away from the faults also suffered
seriously.
According to the Tourism Bureau report, the number of visitors to 230 major scenic
spots dropped by 27 percent for the September-December period of 1999. The room
occupancy rates of hotels for international tourist plummeted by an average of about
60 percent, and international airline reservation cancellations soared to 210,000 at
the same period. Statistics showed that government-operated scenic spots suffered
losses amounting to approximately US$19.5 million, and privately operated tourist
enterprises suffered losses of about US$119 million (Tourism Bureau, 2000).
Huang and Min (2002) examined the recovery of visitor arrivals to Taiwan after the
9-21 Earthquake. The results indicate that Taiwan had not fully recovered from the
sharply reduced inbound arrivals nearly a year after the disastrous earthquake, even
though there was large expenditure on infrastructure rehabilitation. For a long period,
the island had not experienced such an earthquake as the one on September 21.
Since Japan has long been the leading generator of international tourism for
Taiwan, no effort can be spared toward regaining the Japanese tourist market. The
placement of “Peace of Mind” advertisements in major Japanese media was increased,
and promotion seminars were held jointly with large Japanese air carriers in Tokyo,
Fukuoka, Osaka, and Nagoya. Related posters were placed throughout Tokyo railways
and bus stations. Furthermore, a marketing program of “Go, Taiwan” was carried out
in cooperation with China Airlines and Japan Asia Airways. Nevertheless, to attract
tourists to Taiwan, there was still a need to substantially reduce the overall prices of
taking a trip to Taiwan during the early recovery stage. There were signi?cant fare
discounts for Japan-Taiwan routes. Most of the hotels in Taiwan also made many
efforts in response to declining visitor numbers. Many of Taiwan’s international tourist
hotels mailed promotional materials to every previous Japanese guest to try to rescue a
sharply declining Japanese tourist market (Travel Trend News, 1999). In addition to
carrying out measures to revitalize the Japanese tourist market in the aftermath of the
quake, the Taiwan Tourism Bureau also placed 11 advertisements in major Japanese
media during the year 2000 to draw Japanese tourists to visit Taiwan. Moreover, each
Japanese arrival was presented with a box of Alpine Oolong tea while traveling in the
central part of the island to increase the visitation.
IJCTHR
2,3
200
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
2.2 The SARS outbreak, 2003
From the end of 2002 onwards, SARS spread rapidly through the medium of
international travel to more than 26 countries on ?ve continents. Within months of its
?rst appearance in mainland China, SARS had resulted in 8,422 reported probable
cases, of which 916 resulted in deaths. Taiwan had to cope with the third largest SARS
outbreak, with a total of 674 contracted cases and 84 deaths. The largest outbreaks
occurred in Mainland China (5,327 cases and 349 deaths) and Hong Kong (1,755 cases
and 300 deaths) (World Health Organization – WHO, 2003a).
SARS is an acute respiratory illness caused by infection by the corona virus spread
predominantly by droplets and direct or indirect contact. On March 12, 2003, due to the
spread of the illness to several countries in a short period of time, WHO issued a global
alert, suggesting that national authorities implement heightened surveillance for cases
of SARS. Three days later, WHO named the mysterious illness after its symptoms:
SARS, and issued a series of emergency guidelines for travelers and airlines
(WHO, 2003b). The purpose of the announcement was to eliminate the spread of
infection by international travel.
The ?rst case of SARS in Taiwan was a 54-year-old businessman, who had traveled
in Guangdong Province, China, where the ?rst case of SARS is known to have occurred
(WHO, 2003c). The subsequent spread of the epidemic led to WHO designating Taiwan
an affected area on March 15, 2003. Because of the lack of knowledge about the disease,
people had no good information to help them deal with the virus, and tourism
organizations also lacked the resources to protect themselves at home or means of
getting help from abroad. In May, the situation got even worse with the SARS virus
spreading at an alarming rate and the number of new cases and deaths increasing. Fear
and uncertainty blanketed the island. Taipei, the capital of Taiwan, was placed on the
WHO travel advisory list on May 8, and travel warnings were extended to the entire
island on May 21. This was the strictest world-wide travel advisory issued by WHO in
its 55-year history. Following these announcements, 30 nations imposed additional
restrictions on travelers from Taiwan, and up to 48 nations also listed Taiwan as a
travel destination to be avoided (Tourism Bureau, 2003). Consequently, the number of
tourist arrivals fell signi?cantly, reaching levels the island had never witnessed before.
During this time, the Taiwan Government established methods and procedures for
SARS education, prevention and treatment. The effect of the administration’s
decisions, though gradual, began to be seen, and the number of infected cases and
deaths gradually fell. As a result, WHO lifted Taiwan from its SARS travel advisory
list on June 17, 27 days after putting Taiwan on the list, and on July 5 removed it from
the list of areas with recent local transmission.
The SARS epidemic undermined Taiwan’s tourism industry and impeded the ?ow
of tourists. According to a Tourism Bureau report, international arrival numbers fell
sharply in the second quarter by 71.54 percent from a year earlier due to the SARS
outbreak. These declines slowed to 19.73 and 11.1 percent in the third and fourth
quarter, respectively, (Tourism Bureau, 2004). Since Japan is Taiwan’s main source of
visitors, the government and private sector jointly rolled out various tourism
promotion measures and new tour destinations to attract Japanese tourists to the island
after July 5, when WHO removed Taiwan from its list of SARS-affected regions.
Advertising was carried out for tourism promotions through major print and electronic
media channels in Japan. Post-SARS promotions emphasized the safety of travel in the
Forecasting
Japanese tourism
in Taiwan
201
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
island under the themes “Exhausted in Japan, Rest Up in Taiwan,” and “Take it Easy.”
The Taiwan Visitors Association was commissioned to work with domestic tour
operators to form tourism promotional events for the Japanese market. In addition,
media representatives, travel writers and tour operators were invited to take part in
tours from July to September to highlight the safety of traveling in Taiwan after SARS.
The Tourism Bureau sent a delegation to hold promotional events in Tokyo, Osaka
and Nagoya in August to inform the Japanese public about the post-SARS recovery
status of Taiwan. Moreover, several special tour packages were introduced, such as
10,000 yen travel coupon books awarded by draw to 5,000 Japanese travelers. These
swift countermeasures were adopted to revitalize the Japanese market and encourage
Japanese tourism back to the island.
3. Method
3.1 Data collection
Tourist arrivals are de?ned in this study as persons traveling to Taiwan from Japan,
which is not usual residence, for at least 24 hours up to a period not exceeding
12 months. In this research, the Japanese inbound tourism demand series adopted are
unadjusted monthly tourist arrivals from Japan to Taiwan for the period from January
1979 to September 2006, and this set of data has been obtained from the Monthly
Report on Tourism published by the Tourism Bureau of Taiwan. The statistical
package for SAS/ETS 9.1 version software is employed to develop and use the
SARIMA with and without intervention models.
3.2 Time series modeling
Time series analysis can be a valuable tool for forecasters to view trends in data,
both long-term and cyclical. The Box-Jenkins model is the most common time series
technique to be applied. In 1970, Box and Jenkins made the ARIMA model popular and
easy to use by proposing a model building methodology including four stages:
identi?cation, estimation, diagnostic checking and forecasting. This methodology also
allows researchers not only to discover hidden patterns in data but also to generate
forecasts. The model has been providing benchmark forecasts for more than three
decades, and has been fruitfully applied in economics, business, ?nance, engineering,
and many other branches of the physical and social sciences.
The univariate ARIMA model with intervention (also known as intervention
analysis) of a dependent variable is subjected to the effects of exceptional events on a
time series model. It was presented by Box and Tiao (1975) to examine the reduction of
oxidant pollution levels in downtown Los Angeles by adopting two interventions: the
diversion of traf?c by the opening of the Golden State Freeway and the coming into
effect of a new law. Since Box and Tiao ?rst investigated the level shift or drift in the
time series within the framework of the Box-Jenkins transfer function modeling,
intervention analysis has been employed widely to estimate various types of economic
or environmental impacts through policy changes, strikes, promotions, disasters and
the like (Chang and Lin, 1997; Goh and Law, 2002; Kapombe and Colyer, 1999; Lai and
Lu, 2005; Sharma and Khare, 1999; Goh, 2005; Min et al., 2006).
Intervention analysis has been well-documented in modeling and forecasting
tourism demand. Bonham and Gangnes (1996) adopted time series intervention
analysis to evaluate ex post the effect of the 1987 Hawaii hotel room tax imposition.
IJCTHR
2,3
202
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Empirical ?ndings show that there was no statistically signi?cant evidence of any
impact by the tax. Lee et al. (2005) used the intervention model to estimate the impact of
the September 11 terrorist attacks on US air transport passenger demand and to see
how long it took the US air transport industry to recover from the crisis. Coshall (2003)
applied the intervention model to assessing the impact of three interventions, the US
bombing of Libya in 1986, the Lockerbie air disaster in 1988, and the Persian Gulf crisis
during 1990-1991, on the ?ow of UK air passengers to a variety of destinations. In
another study (Coshall, 2005), used the intervention approach to measure the impact of
different events, namely terrorism, war, and the foot and mouth disease outbreak, on
expenditure by visitors to the UK and UK tourists abroad. Min et al. (2006) used a
SARIMA with intervention model both to evaluate the impact of the SARS outbreak in
Taiwan and incorporated the intervention into the outbound time series model to
improve parameter estimates and forecasts. Studies by Goh and Law (2002) and Lai
and Lu (2005), compared the SARIMA with intervention model with different
techniques; the results showed that it outperformed all other models when signi?cant
intervention in the series existed. Based on this literature review, the following
hypothesis is tested in this study:
H1. Intervention analysis of Japanese tourism demand in Taiwan after the
9-21 Earthquake and the SARS outbreak produces better forecasts in
comparison with forecasts without intervention analysis.
3.3 Mathematical formulation
In univariate ARIMA analysis, the time series to be modeled is based on observations
generated by an underlying process to be identi?ed. In addition, one of the basic
principles of model building is that the model should adequately represent the data
using as few parameters as possible (Box and Jenkins, 1970). This section is a brief
review formulation of the time series model.
The goal is to ?nd a good model that adequately represents the process-generating
mechanism. There are two useful representations to express the behavior of observed
time series processes, namely the autoregressive (AR) and moving average (MA)
models. The AR model is used to describe a time series in which the current
observation depends on its past values, whereas the MA model is used to describe a
time series process as a linear function of current and past random shocks. The sample
partial autocorrelation function and the sample autocorrelation function (ACF) are
adopted to identify the models in AR and MA, respectively. The general formulation
suggested by Box and Jenkins as an autoregressive moving average can be written as:
fðBÞZ
t
¼ uðBÞa
t
ð1Þ
where Z
t
and a
t
represent the Japanese inbound numbers and random error terms at
time t, respectively. B is a backward shift operator de?ned by BZ
t
¼ Z
t21
. f(B) and
u(B) are AR and MA operators of orders p and q, respectively, and are expressed as:
fðBÞ ¼ 1 2f
1
B 2f
1
B 2f
3
B 2. . . 2f
p
B
p
ð2Þ
and:
uðBÞ ¼ 1 2u
1
B 2u
2
B 2u
3
B 2· · · 2u
q
B
q
ð3Þ
Forecasting
Japanese tourism
in Taiwan
203
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
If a time series presents nonstationary, it can often be transformed into a stationary
series by taking ?rst difference. Nonstationary models after differencing can be
expressed as:
fðBÞð1 2BÞ
d
Z
t
¼ uðBÞa
t
ð4Þ
(1 2 B)
d
Z
t
means d of difference of the process to be stationary, which also can be
written as ARIMA( p, d, q).
Tourist arrivals series often exhibits a seasonal periodic component which
measures at regular calendar intervals within a year. Box and Jenkins have generalized
the ARIMA model to deal with seasonality and de?ne it as SARIMA. It is denoted as
ARIMA( p, d, q )x(P,D,Q)
s
in which p and P are the orders of the AR operator; d and D
are the differencing; and q and Qare the orders of the MA of non-seasonal and seasonal
components, respectively. It combines the ARIMA( p, d, q ) and yields:
fðBÞFðB
s
Þð1 2BÞ
d
ð1 2B
s
Þ
D
Z
t
¼ uðBÞQðB
s
Þa
t
ð5Þ
For intervention analysis, it is to study the impact of exceptional external events as:
y
t
¼
X
k
i¼1
v
i
ðBÞ
d
i
ðBÞ
B
b
i
j
ti
þ N
t
ð6Þ
where v
i
ðBÞ ¼ v
i0
2v
i1
B 2· · · 2v
is
i
B
s
i
and d
i
ðBÞ ¼ 1 2d
i1
B 2· · · 2d
ir
i
B
r
i
are the
numerator and denominator polynomial of the intervention effects function,
respectively. j
ti
is represented by a series of 0’s and 1’s denoting non-occurrence and
occurrence of the intervention. B is the backshift operator, de?ned as
By
t
¼ y
t21
and B
b
y
t
¼ y
t2b
. N
t
is the noise series which represents the background
observed series y
t
without the intervention effects.
3.4 Types of intervention responses
Box and Tiao (1975) considers the intervention variables as having a pulse or step
function. If the effects of an intervention are temporary or transient and will die out
after time t, a pulse function is used, where:
j
t
¼ P
ðTÞ
t
¼
0
1
(
t – T
t ¼ T
ð7Þ
If the effects of an intervention are expected to remain permanently after time t to some
extent, a step function is adopted, where:
j
t
¼ S
ðTÞ
t
¼
0
1
(
t , T
t $ T
ð8Þ
In addition, Box and Tiao illustrated some typical transfer functions for step and pulse
response patterns shown in Figure 2. The following Figure 2(a) can be used as a step or
abrupt permanent effect with level of magnitudev. Figure 2(b) shows the gradual
permanent effect with change rate d and long-run response level v/1 2 d. Figure 2(c) is
a special case of Figure 2(b) as d ¼ 1. Figure 2(d) shows an abrupt and temporary effect
IJCTHR
2,3
204
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
with a gradual decay of rate d that will approach the pre-intervention level with no
permanent effect. Figure 2(e) is similar to Figure 2(d), but with a permanent effect of
magnitude v
2
. Figure 2(e) is a linear combination of Figure 2(a) with Figure 2(d), the
pattern of Figure 2(a) is equivalent to ðvB=1 2BÞP
ðTÞ
t
through the relationship
ð1 2BÞS
ðTÞ
t
¼ P
ðTÞ
t
. Figure 2(f) can be used instead of Figure 2(e) if an initial impact is
considered.
4. Empirical results
Intervention analysis begins by identifying a plausible set of ARIMA models.
Generally, the longest data span of the pre- and post-intervention observations is used
to identify N
t
(Ender, 2004; Coshall, 2003). Therefore, the data adopted in this study
consist of monthly visitor arrivals from Japan during successive months over a 27-year
period ( January 1979-September 2006). Among the 333 observations, the ?rst
321 observations ( January 1979-September 2005) are used to develop two tentative
models, with and without intervention analyses. These two models are employed to
Figure 2.
Forms of intervention
responses
w B S
t
(T)
STEP:
w
i
(B)
d
i
(B)
S
t
(T)
PULSE:
w
i
(B)
d
i
(B)
P
t
(T)
w B
1 – d (B)
S
t
(T)
w B
1 – d (B)
P
t
(T)
w B
1 – B
S
t
(T)
w
w
1
w
1
w
0
w
1
w
2
w
2
w
1–d
w
(b)
(c)
(a) (d)
(e)
(f)
w
1
B
1 – d B
P
t
(T)
w
2
B
1 – B
+
w
1
B
w
0
1 – d B
P
t
(T)
w
2
B
1 – B
+
Forecasting
Japanese tourism
in Taiwan
205
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
forecast the 12 future values of Japanese inbound arrivals of October 2005-September
2006, and then compare with the known values for accuracy testing.
The ?rst aspect of a time series analysis is to plot the data; Figure 3 shows this
point. The plot exhibits some interesting features. First, an attention is immediately
drawn to an obvious decline from the 292nd observation (April, 2003), which is caused
by the SARS outbreak. Another clear reduction can also be noticeably identi?ed from
the 248th observation (September, 1999), relates to the 9-21 Earthquake. Second, an
initial plot of the data of Japanese visitor arrivals reveals irregular variation and
seasonal variation. The two disasters in this study, the 9-21 Earthquake and the SARS
epidemic, have a profound in?uence on the Japanese tourism demand for travel to
Taiwan, and both are intervention events in the Taiwan’s tourism.
4.1 Pre-intervention model
The appropriate model for noise component N
t
is identi?ed prior to the known
interventions, the 9-21 Earthquake and SARS, and the ?rst 248 observations, January
1979 through August 1999, are utilized for the identi?cation process. There is a
question of whether the series is stationary or not. The time series plot of the ?rst
248 observations is clearly nonstationary. To con?rm this, the ACF and augmented
Dickey-Fuller test (ADF) are used to test for stationarity in the available time-series
data. For the ADF test, the null hypothesis is that the series are nonstationary and
therefore include a unit root. The representations and hypothesis of the ADF test are as
follows:
Dy
t
¼ b
0
þ dt þ b
1
y
t21
þ
X
p
i¼1
g
i
Dy
t2i
þ 1
t
ð9Þ
An ADF regression investigates the null hypothesis of non-stationarity (b
1
¼ 0)
against the alternative of stationarity (b
1
, 0).
The time series represents nonstationary by the ADF test. ACF also shows positive
and high for a number of lags; and decays fairly slowly. It indicates that the ?rst
248 time series is nonstationary. Thus, differencing is necessary. However, the use of
(1 2 B) alone does not remove the effects of nonstationarity from the data, and the ACF
Figure 3.
Time series plot on
Japanese visitor arrivals
from January 1979 to
September 2006
0
20,000
40,000
60,000
80,000
1,00,000
1,20,000
1
9
7
9
1
9
8
1
1
9
8
3
1
9
8
5
1
9
8
7
1
9
8
9
1
9
9
1
1
9
9
3
1
9
9
5
1
9
9
7
1
9
9
9
2
0
0
1
2
0
0
3
2
0
0
5
Year
N
u
m
b
e
r
o
f
J
a
p
a
n
e
s
e
a
r
r
i
v
a
l
s
Intervention Events
IJCTHR
2,3
206
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
at lags 12, 24, 36, and 48 shows the most distinct seasonal effect, indicating the
presence of seaonality in the given time series. Hence, seasonal differencing is
warranted. In view of the above indications, the ?rst and seasonal difference
(1 2 B)(1 2 B
12
), should be taken in order to achieve stationariy. The pattern is
indicative of a multiplicative MA(1) and MA(12) model, that is (1 2 u
1
B)(1 2 u
12
B
12
)
after the ACF diagrams and ADF tests, and the Ljnug-Box Q statistic for residual
checking. Following the identi?cation, the estimate for the Japanese tourist of the
pre-intervention is yielded by Maximum Likelihood Estimation as:
ð1 2BÞð1 2B
12
ÞZ
t
¼ ð1 20:54451BÞð1 20:60876B
12
Þa
t
ð10Þ
4.2 SARIMA model with intervention
The Japanese inbound series in this study indicates that the ?rst impact of the
9-21 Earthquake and SARS interventions starts to take effect in September 1999 and
April 2003, respectively. The intervention variables j
t1
and j
t2
are set to value 1 for
September 1999 and April 2003 and 0 otherwise. According to Figure 3, the effect of the
single impulse lasts more than one month due to the SARS impact, a second drop in
May 2003 suggests a numerator lag and the following exponential increase suggests
a denominator lag in the transform function (Brocklebank and Dickey, 2003). As the
impact of the earthquake is illustrated as an abrupt, a temporary impact (shown
in Figure 3), we expect the 9-21 Earthquake has a similar type of intervention
as the SARS incident. Therefore, the two intervention models are set in the form
ðv
o1
2v
11
B=1 2d
11
BÞj
t1
and ðv
o2
2v
12
B=1 2d
12
BÞj
t2
.
The Maximum Likelihood estimates of the SARIMA model with intervention are:
Z
t
¼
v
o1
2v
11
B
1 2d
11
B
j
t1
þ
v
o2
2v
12
B
1 2d
12
B
j
t2
þ
ð1 2u
1
BÞð1 2u
12
B
12
Þ
ð1 2BÞð1 2B
12
Þ
a
t
Z
t
¼
20:20654 20:44164B
1 20:80722B
j
t1
þ
20:64039 21:95801B
1 20:63207B
j
t2
þ
ð1 20:60420BÞð1 20:65753B
12
Þ
ð1 2BÞð1 2B
12
Þ
a
t
ð11Þ
Since Dubin –Watson test is designed to detect ?rst order AR errors, it is therefore
used to examine if autocorrelation is present. It can be expressed as:
d ¼
X
n
t¼2
^ 1
t
2 ^ 1
t21
ð Þ
2
=
X
n
t¼1
^ 1
2
t
ð12Þ
where:
^ 1
t
¼ Y
t
2
^
b
0
2
^
b
1
X
1t
2
^
b
2
X
2t
ð13Þ
If the actual errors e
t
are uncorrelated, the numerator of d has an expected value of
approximately2(n 2 1)s
2
andthe denominator has anexpectedvalue of about ns
2
. Thus,
with errors uncorrelated, the ratio d should be about 2 (Brocklebank and Dickey, 2003).
The result of Dubin-Watson test in the intervention model obtains 2.0054, indicating the
errors are uncorrelated.
Forecasting
Japanese tourism
in Taiwan
207
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Table I presents the results of the estimation and shows that all estimate parameters
are signi?cant with very small p-values. The plot of the estimated residual ACFis shown
in Figure 4, and it is evident fromthe ?gure that the residuals are not autocorrelated, and
reveal approximate white noise. Moreover, the Ljung-Box Q statistic is not signi?cant
and indicates the SARIMA model with intervention does ?t well.
Parameter Estimate Standard error t-value Approx Pr . jtj Lag Variable
u
1
0.60420 0.04496 13.44 ,0.0001 1 MA(1)
u
12
0.65753 0.04490 14.64 ,0.0001 12 MA(12)
v
o1
20.20654 0.08581 22.41 0.0161 0 9-21Earthquake
v
11
0.44164 0.09048 4.88 ,0.0001 1 9-21Earthquake
d
11
0.80722 0.05920 13.64 ,0.0001 1 9-21Earthquake
v
o2
20.64039 0.13916 24.60 ,0.0001 0 SARS
v
12
1.95801 0.09058 21.62 ,0.0001 1 SARS
d
12
0.63207 0.02963 21.33 ,0.0001 1 SARS
Table I.
Full-intervention model
by maximum likelihood
estimation
Figure 4.
Estimated residual ACF
on ARIMA model with
intervention
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
–1 –0.5 0 0.5 1
Correlation coefficients
Autocorrelations
IJCTHR
2,3
208
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
The permanent effect of the earthquake impact can be investigated by adding one level
shift variable L
t
, and de?ning the level shift variable value to be 1 prior to April 2003
and 0 otherwise. The coef?cient of permanent effect l
2
¼ 0.26047, with standard error
0.13537, is not signi?cant. This means the permanent effect is not signi?cantly
different from 0, indicating no permanent effect after the SARS event. This is
equivalent to the term {v
1
B=1 2B}P
ðTÞ
t
, with permanent effect v
1
, and the transfer
function suggested by Box and Tiao (1975) as following:
v
0
þ
v
1
B
1 2dB
þ
v
2
B
1 2B
P
ðTÞ
t
ð14Þ
The in?uence of change rate d in event of SARS, impacting the effect of transfer
function I
t
of the Japanese inbound series, is considered in the intervention model
as follows:
I
t
¼
v
o2
2v
12
B
1 2d
12
B
j
t
¼ ðv
o2
2v
12
BÞð1 þd
12
B þd
2
12
B
2
þd
3
12
B
3
þ . . .Þj
t
¼ v
o2
j
t
þ ðv
o2
d
12
2v
12
Þj
t21
þd
12
ðv
o2
d
12
2v
12
Þj
t22
þ d
2
12
ðv
o2
d
12
2v
12
Þj
t23
þ . . .
ð15Þ
The same procedure to test the permanent effect of the earthquake is also applied.
The results show that the coef?cient of permanent effect l
1
¼ 20.08413, with
standard error 0.22510, is not signi?cant either, indicating no permanent effect after the
earthquake.
4.3 ARIMA model
As a ?rst step to model identi?cation, the plot of the data for the monthly Japanese
tourist arrivals from January 1979 to September 2005 is considered. The series is
checked for stationarity using ACF and ADF for unit roots and, if necessary, the series
is transformed by taking appropriate differences to render the series stationary. The
t-statistics of ADF test on the series at level are larger than the critical value at the
5 percent signi?cance level. Thus, the null hypothesis of unit root is not rejected,
indicating the series is nonstationary. Differencing is necessary. Since the series exhibits
varying seasonal patterns, it is essential to test for the presence of seasonal unit roots.
After seasonal difference (1 2 B
12
), the series achieves stationarity. Various SARIMA
models are ?tted to the Japanese arrivals series after seasonal differencing. Since the
Akaike information criterion (AIC) and the Schwarz Bayesian criterion (SBC) are
generally used in model selection criteria, whereby the smaller values are preferred;
thus, the selected models are then compared according to the AIC and SBC. These
procedures are repeated until anappropriate model is found. After best model is selected,
the Dubin-Watson test is adopted to obtain 1.9994, demonstrating the errors are
uncorrelated. Furthermore, residual checking using p-value associated with the
Ljnug-Box Qstatistic shows that the residual of the model represents random, indicating
the residuals approximate white noise. The ACF of the residual series is also used to
check residuals (Figure 5). The ACF is clean and reveals potential existent white noise
and no anomalies. Following the identi?cation, the estimate for the SARIMA without
intervention model is obtained by Maximum Likelihood Estimation as:
Forecasting
Japanese tourism
in Taiwan
209
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
ð1 20:67361BÞð1 2B
12
ÞZ
t
¼ ð1 þ 0:36322BÞð1 20:73315B
12
Þa
t
ð16Þ
4.4 Evaluation of two models
The two models are estimated using the data from January 1979 to September 2005.
Accuracy measurement of the two approaches for the period October 2005 to
September 2006 is based on mean absolute deviation (MAD), mean absolute percentage
error (MAPE) and mean square error (MSE), which are often adopted for the
performance of each model. Each of these measurements is de?ned as follows:
MAD ¼
1
n
X
n
i¼1
Diff
i
j j ð17Þ
MAPE ¼
1
n
X
n
i¼1
Diff
i
j j
Act
i
£ 100 percent ð18Þ
Figure 5.
Estimated residual
ACF on ARIMA model
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
–1 –0.5 0 0.5 1
Correlation coefficients
Autocorrelations IJCTHR
2,3
210
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
MSE ¼
1
n
X
n
i¼1
Diff
2
i
ð19Þ
Table II shows the summary of SARIMA and intervention models evaluation, and
Figure 6 shows a graphical presentation of the values generated by the two forecasting
methods.
As Table II shows, the forecasting output from the intervention analysis is more
accurate, with a relatively small amount of error. The minimum values of MAD, MAPE
SARIMA SARIMA with intervention
Variable (1,0,1)(0,1,1) (0,1,1)(0,1,1)
MA(1) u
1
20.363
* *
0.571
* *
MA(12) u
12
0.733
* *
0.660
* *
AR(1) f
1
0.674
* *
9-21earthquake v
o1
20.209
*
9-21earthquake v
11
0.435
* *
9-21earthquake 0.812
* *
SARS v
o2
20.865
* *
SARS v
12
1.983
* *
SARS d
12
0.663
* *
AIC 2273.3 2532.6
SBC 2262.1 2502.8
Residuals Q(36) 0.888 0.220
R
2
0.807 0.785
MAD(2005/10-2006/09) 0.2302 0.0561
MAPE(2005/10-2006/09) 2.01 percent 0.49 percent
MSE(2005/10-2006/09) 0.0671 0.0052
Notes: Signi?cant levels at:
*
0.05 and
* *
0.01, respectively
Table II.
Evaluation of SARIMA
and intervention models
Figure 6.
Graphical presentation
of two forecasting
methods
Actural visitors (1,0,1)(0,1,1) (0,1,1)(0,1,1)with intervention
50,000
60,000
70,000
80,000
90,000
1,00,000
1,10,000
1,20,000
Forecasting month
N
u
m
b
e
r
o
f
J
a
p
a
n
e
s
e
v
i
s
i
t
o
r
s
t
o
T
a
i
w
a
n
1 2 3 4 5 6 7 8 9 10 11 12
Forecasting
Japanese tourism
in Taiwan
211
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
and MSE show that the deviations between the predicted values derived by the
intervention analysis and the actual values are smaller. The R
2
of the intervention
model is also larger. In addition, based on the criterions of AIC and SBC,
the intervention model is better than the model of SARIMA without intervention.
As Figure 6 shows, the forecast data from the intervention model are closer to the
actual data. In view of the above-empirical examinations, intervention model
outperforms the SARIMA without intervention in describing the Japanese tourist ?ows
after the events of the 9-21 Earthquake and SARS. The statistical results not only
support the hypothesis, but are also consistent with previous studies on intervention
analysis in the tourism literature.
5. Implications and conclusions
This study adopts an intervention model to estimate the impacts of two disasters, the
9-21 Earthquake and SARS, and to explore if they had a temporary or long-term impact
on Japanese inbound tourism demand. The intervention analysis is compared to a
model without intervention to see which technique can accurately predict Japanese
tourist ?ows to Taiwan under the in?uence of a drastic environmental change such as
an earthquake or epidemic. The forecasting performance is evaluated using
out-of-sample predictions. According to the ?ndings, the effect of both disasters on
Japanese inbound tourism presents only temporarily. Experimental results also
indicate that the forecasting ef?ciency of ARIMA with intervention is superior to that
of a model without intervention.
The results indicate that the Japanese inbound arrivals sharply dropped following
both of the two disastrous occurrences, suggesting that the Japanese tourists are likely
to be responsive to prompt marketing strategies and messages. Since Japanese tourists
are more adverse to risky leisure pursuits (Polunin, 1989; Ahmed and Krohn, 1992), the
safety of accommodations and recreational facilities, and the security of the
environment should be strongly advertised to help the Japanese minimize the feeling of
uncertainty. By doing this, the Japanese tourists can be overcome without diluting
interest in traveling to Taiwan. Moreover, the propensity of package tour travel is
strong among Japanese tourists (Lim and McAleer, 2005). The practical implication for
tourism operators include the usefulness of reinforcing the package holiday by
establishing an attractively priced travel package or offering a package with a variety
of highly desirable or unique features to increase competition. Furthermore, because
Japanese tourists are more at ease in groups (Ahmed and Krohn, 1992; Pinhey and
Iverson, 1994), tour packages should promote group activities rather than organizing
activities for them in isolation from other people, in order to ful?ll their need for
security. Printed advertisements and packages offered through travel agents could
emphasize the fact that tourism destinations in Taiwan are still functional in order to
rekindle Japanese visitors’ interest.
This study is a ?rst attempt in the tourism literature to model Japanese demand for
travel to Taiwan after these two traumatic crises. However, there are some limitations
that could be overcome in future research. This study had dif?culty accurately
delineating the rebound in Japanese tourist based on monthly data. It may be more
accurate to classify recovery periods using weekly data in future studies. In addition,
tourismresearchers can apply other forecasting techniques to produce optimal results in
studies of the effects of unexpected misfortunes in the future. Since arti?cial neural
IJCTHR
2,3
212
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
networks are able to outperformothers in most situations (Law, 2001; Burger et al., 2001),
these can be used to investigate which generate more accurate forecasts. Also, a possible
future study would be to test the accuracy of a vector autoregressive model capable of
producing accurate medium to long-term forecasts, which does not require the
generation of forecasts for explanatory variables before forecasts of dependent variables
can be obtained (Song and Witt, 2006). Another suggested direction for further research
is the ARMAX modeling approach, which is a single equation econometric technique
that accounts for the dynamic relationship between variables such as GDP and
exchange rates. Accurate predictions from such forecasting techniques can aid tourist
authorities and professionals in Taiwan improve their planning and decision-making
processes. Lastly, there are other factors that might in?uence a rebound, such as people’
fading memories or the purpose of visitation. The geographic proximity of Taiwan to
Japan could also account for perceived risk factors. Whether these aspects represent
unexamined factors in this study could become the basis of further studies.
The value of forecasting lies in its ability to assist tourism authorities make
operational, tactical and strategic decisions (Wang and Lim, 2005; Law and Au, 1999).
Therefore, in deciding which forecasting method to use after a negative occurrence,
practitioners and policy makers may con?dently apply intervention analysis.
Although both these events impeded the ?ow of Japanese tourists, traf?c almost
returned to pre-catastrophe levels afterwards. Thus, examining such effects in a
detailed and accurate manner enables tourism authorities to learn and appreciate the
dynamics of tourist behavior for future negative occurrences to achieve more
sophisticated forward planning. This research will potentially bene?t researchers of
tourism and tourism authorities in Taiwan as well as worldwide.
References
Ahmed, Z.U. and Krohn, F.B. (1992), “Understanding the unique consumer behavior of Japanese
tourists”, J. of Travel & Tourism Market, Vol. 3 No. 1, pp. 73-86.
Balaz, V. and Mitsultake, M. (1998), “Japanese tourists in transition countries of Central Europe:
present behavior and future trends”, Tourism Manage, Vol. 19 No. 5, pp. 433-43.
Bonham, C.S. and Gangnes, B. (1996), “Intervention analysis with cointegrated time series:
the case of the Hawaii hotel room tax”, Appl. Econ., Vol. 28 No. 10, pp. 1281-93.
Box, G.E.P. and Jenkins, G.M. (1970), Time Series Analysis: Forecasting and Control, Holden-Day
Inc., San Francisco, CA.
Box, G.E.P. and Tiao, G.C. (1975), “Intervention analysis with applications to economic and
environmental problems”, J. of the Am. Stat. Associ., Vol. 70, pp. 70-4.
Brocklebank, J. and Dickey, D.A. (2003), SAS for Forecasting Time Series, SAS Institute Inc.,
Cary, NC.
Burger, C.J.S.C., Dohnal, M., Kathrada, M. and Law, R. (2001), “A practitioners guide to
time-series methods for tourism demand forecasting – a case study of Durban,
South Africa”, Tourism Manage, Vol. 22 No. 4, pp. 403-9.
Cha, S., McCleary, K.W. and Uysal, M. (1995), “Travel motivations of Japanese overseas travelers:
a factor-cluster segmentation approach”, J. of Travel. Res., Vol. 34 No. 1, pp. 33-9.
Chang, N. and Lin, Y.T. (1997), “An analysis of recycling impacts on solid waste generation by
time series intervention modeling”, Resour, Cons. and Recycling, Vol. 19 No. 3, pp. 165-85.
Forecasting
Japanese tourism
in Taiwan
213
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Chen, J.S. (2000), “Cross-cultural differences in travel information acquisition among tourists
from three Paci?c-rim countries”, J. of Hospitality &TourismRes., Vol. 24 No. 2, pp. 239-51.
Choy, D.J.L. (1998), “Changing trends in Asia-Paci?c tourism”, Tourism Manage, Vol. 19 No. 4,
pp. 381-2.
Coshall, J. (2003), “The threat of terrorism as an intervention on international travel ?ows”, J. of
Travel. Res., Vol. 42 No. 1, pp. 4-12.
Coshall, J. (2005), “Interventions on UK earnings and expenditures overseas”, Ann. of Tourism
Res., Vol. 32 No. 5, pp. 592-609.
Dybka, J. (1988), “Overseas travel to Canada: new research on the perceptions and preferences of
the pleasure travel market”, J. of Travel Res., Vol. 26 No. 4, pp. 12-15.
Ender, W. (2004), Applied Econometrics Time Series, Willey, New York, NY.
Faulkner, B. (2001), “Towards a framework for tourism disaster management”, Tourism
Manage, Vol. 22 No. 2, pp. 135-47.
Goh, B.H. (2005), “The dynamic effects of the Asian ?nancial crisis on construction demand and
tender price levels in Singapore”, Building and Environ, Vol. 40 No. 2, pp. 267-76.
Goh, C. and Law, R. (2002), “Modeling and forecasting tourism demand for arrivals with
stochastic nonstationary seasonality and intervention”, Tourism Manage, Vol. 23 No. 5,
pp. 499-510.
Heung, V.C.S., Qu, H. and Chu, R. (2001), “The relationship between vacation factors and
socio-demographic and traveling characteristics: the case of Japanese leisure travelers”,
Tourism Manage, Vol. 22 No. 3, pp. 259-69.
Holtzman, W.L., Murthy, B. and Gordon, J.C. (1991), “Cultural bridging with the Japanese”,
Cornel Hotel and Rest Admin Quart, Vol. 32 No. 3, pp. 52-9.
Huang, J.H. and Min, J.C.H. (2002), “Earthquake devastation and recovery in tourism: the Taiwan
case”, Tourism Manage, Vol. 23 No. 2, pp. 145-54.
Iverson, T.J. (1997a), “Decision timing: a comparison of Korean and Japanese travelers”, Int J. of
Hospitality Manage, Vol. 16 No. 2, pp. 209-19.
Iverson, T.J. (1997b), “Japanese visitors to Guam: lessons from experience”, J. of Travel &
Tourism Market, Vol. 6 No. 1, pp. 41-54.
Kapombe, C.M. and Colyer, D.A. (1999), “Structural time series analysis of US broiler exports”,
Agri. Econ., Vol. 21 No. 3, pp. 295-307.
Lai, S.L. and Lu, W.L. (2005), “Impact analysis of September 11 on air travel demand in the USA”,
J. of Air Trans. Manage, Vol. 11 No. 6, pp. 455-8.
Lang, C.T., O’Leary, J.T. and Morrison, A.M. (1993), “Activity segmentation of Japanese female
overseas travelers”, J. of Travel & Tourism Market, Vol. 2 No. 4, pp. 1-22.
Law, R. (2001), “The impact of the Asian ?nancial crisis on Japanese demand for travel to
Hong Kong: a study of various forecasting techniques”, J. of Travel & Tourism Market,
Vol. 10 Nos 2/3, pp. 47-65.
Law, R. and Au, N. (1999), “A neural network model to forecast Japanese demand for travel to
Hong Kong”, Tourism Manage, Vol. 20 No. 1, pp. 89-97.
Lee, S., Oh, C. and O’Leary, J.T. (2005), “Estimating the impact of the September 11 terrorist
attacks on the US air transport passenger demand using intervention analysis”, Tourism
Anal., Vol. 9 No. 4, pp. 355-61.
Lim, C. and McAleer, M. (2005), “Analyzing the behavioral trends in tourist arrivals from Japan
to Australia”, J. of Travel. Res., Vol. 43 No. 4, pp. 414-21.
IJCTHR
2,3
214
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Lim, C. and Wang, Y. (2005), “Forecasting Japanese outbound tourism using seasonal time series
models”, Int. J. of Hospitlaiy Manage., Vol. 3 No. 2, pp. 37-54.
Lin, M. (1990), “Taiwan: magnet for Japanese tourists”, Cornel Hotel and Rest Admin Quart,
Vol. 31 No. 1, pp. 96-7.
Mak, J., Carlile, L. and Dai, S. (2005), “Impact of population aging on Japanese international travel
to 2025”, J. of Travel. Res., Vol. 44 No. 2, pp. 151-76.
March, R. (1997a), “An exploratory study of buyer-supplier relationships in international
tourism: the case of Japanese wholesalers and Australian suppliers”, J. of Travel &
Tourism Market, Vol. 6 No. 1, pp. 55-68.
March, R. (1997b), “Diversity in Asian outbound travel industries: a comparison between
Indonesia, Thailand, Taiwan, South Korea and Japan”, Int. J. of Hospitality Manage.,
Vol. 16 No. 2, pp. 231-8.
Mihalik, B.J., Uysal, M. and Pan, M.C. (1993), “A comparison of information sources used by
vacationing Germans and Japanese”, Hospitality Res. J., Vol. 18 No. 3, pp. 39-46.
Min, J., Wu, B. and Wu, K. (2006), “Forecasting outbound tourism with intervention analysis”,
The Bus Review, Vol. 6 No. 2, pp. 179-86.
Murakami, K. and Go, F. (1990), “Transitional corporations capture Japanese market”, Tourism
Manage, Vol. 11 No. 4, pp. 348-52.
Murphy, A. and Williams, P.W. (1999), “Attracting Japanese tourists into the rural hinterland:
implications for rural development and planning”, Tourism Manage, Vol. 20 No. 4,
pp. 487-99.
Nozawa, H.A. (1992), “Marketing analysis of Japanese outbound travel”, Tourism Manage,
Vol. 13 No. 2, pp. 226-33.
Pinhey, T.K. and Iverson, T.J. (1994), “Safety concerns of Japanese visitors to Guam”, J. of Travel
& Tourism Market, Vol. 3 No. 2, pp. 87-94.
Polunin, I. (1989), “Japanese travel boom”, Tourism Manage, Vol. 10 No. 1, pp. 4-8.
Sage, C. (1985), “Concerning the Japanese market”, Tourism Manage, Vol. 7 No. 4, pp. 301-3.
Sakai, M., Brown, J. and Mak, J. (2000), “Population aging and Japanese international travel in the
21st century”, J. of Travel Res., Vol. 38 No. 3, pp. 212-20.
Sharma, P. and Khare, M. (1999), “Application of intervention analysis for assessing the
effectiveness of CO pollution control legislation in India”, Trans. Res. Part D, Vol. 4 No. 6,
pp. 427-32.
Song, H. and Witt, S.F. (2006), “Forecasting international tourist ?ows to Macau”, Tourism
Manage, Vol. 7 No. 2, pp. 214-24.
Tokuhisa, T. (1980), “Tourism within, from and to Japan”, Int. Social Sci. J., Vol. 32 No. 1,
pp. 128-50.
Tourism Bureau (2000), 1999 Annual Report on Tourism, Tourism Bureau, Taipei.
Tourism Bureau (2003), “The status of tourism industry in?uenced by SARS”, available at:http://anonymouse.ws/cgi-bin/anon-www.cgi/http:/202.39.225.133/sars/c04.htm
Tourism Bureau (2004), 2003 Annual Report on Tourism, Tourism Bureau, Taipei.
Travel Trend News (1999), “1999 special report: marketing strategies for inbound and outbound
markets after the 9-21 Earthquake”, Travel Trend News, Vol. 209, pp. 16-27.
Wang, Y. and Lim, C. (2005), “Assessing forecasting models for outbound Japanese travellers to
Australia and China”, J. of Appl. Econ. and Manage., Vol. 3 No. 2, pp. 1-15.
Forecasting
Japanese tourism
in Taiwan
215
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
WHO (2003a), “Summary table of SARS cases by country”, World Health Organization,
November 1, 2002-August 7, 2003, available at: www.who.int/csr/sars/country/2003_08_
15/en/f (accessed August 15, 2006).
WHO (2003b), “World Health Organization issues emergency travel advisory”, World Health
Organization, available at: www.who.int/csr/sars/archive/2003_03_15/en/ (accessed
March 15, 2006).
WHO (2003c), “Taiwan, China: transmission interrupted in last outbreak area”, World Health
Organization, available at: www.who.int/csr/don/2003_07_05/en/ (accessed July 5, 2006).
World Tourism Organization (2006), “UNWTO world tourism barometer”, available at: www.
unwto.org/facts/menu.html (accessed September 6, 2006).
Corresponding author
Jennifer C.H. Min can be contacted at: [email protected]
IJCTHR
2,3
216
To purchase reprints of this article please e-mail: [email protected]
Or visit our web site for further details: www.emeraldinsight.com/reprints
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
This article has been cited by:
1. Tianshu Zheng, John Farrish, Miranda Kitterlin. 2016. Performance Trends of Hotels and Casino Hotels
Through the Recession: An ARIMA With Intervention Analysis of Stock Indices. Journal of Hospitality
Marketing & Management 25, 49-68. [CrossRef]
2. Bo Peng, Haiyan Song, Geoffrey I. Crouch. 2014. A meta-analysis of international tourism demand
forecasting and implications for practice. Tourism Management 45, 181-193. [CrossRef]
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
2
:
0
5
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
doc_325107960.pdf