Content uploaded by Yang Yang
Author content
All content in this area was uploaded by Yang Yang on Feb 26, 2017
Content may be subject to copyright.
1
DETERMINANTS OF LENGTH OF STAY FOR DOMESTIC
TOURISTS: CASE STUDY OF YIXING
Yang Yang
University of Florida
Kevin. K.F Wong
The Hong Kong Polytechnic University
Jie Zhang
Nanjing University
Yang Yang, (Department of Geography, University of Florida, Gainesville, FL,
32611, U.S.A, Email: yang.yang@ufl.edu, phone: 13528716187), is a Ph.D student
in Department of Geography, University of Florida.
Kevin K.F Wong, PhD (Email: hmkevinw@inet.polyu.edu.hk, phone:
85227666341), is an associate professor in the School of Hotel and Tourism
Management, the Hong Kong Polytechnic University. His address is School of Hotel
and Tourism Management, the Hong Kong Polytechnic University, Hung Hom,
Kowloon, Hong Kong.
Jie Zhang, PhD (email: jiezhang@nju.edu.cn), is a professor in Institute of
Tourism Studies, Nanjing University, PRC.
Yang, Y., Wong, K.and Zhang, J. (2011). Determinants of length of stay
for domestic tourists: Case study of Yixing. Asia Pacific Journal of
Tourism Research, 16(6), 619633.
2
DETERMINANTS OF LENGTH
OF STAY FOR DOMESTIC TOURISTS:
CASE STUDY OF YIXING
The length of stay of a tourist is one of the most important factors indicating consumption levels
and revenue generation for certain tourist destinations. This study employs data from a tourist
survey in Yixing, China to investigate potential factors influencing a tourist’s length of stay.
Applying a (generalized) ordered logit model, it is found that distance, age, organized tour,
transportation, motivation, past visits, and assessment of accommodation are some of the major
determinants of a tourist’s length of stay. The results indicate that traveling distance and the
assessment of accommodation are positively associated with the length of stay. In addition,
tourists with different modes of transportation, motivations, and past visits have different
durations of stay. Based on the estimation results from subsamples, it is also found that there are
differences in determinants of length of stay between organized tourists and individual tourists,
and among different age groups.
KEYWORDS: length of stay; generalized ordered logit model; Yixing
INTRODUCTION
In the tourism industry, one of the most important indices in evaluating tourists’ demands and
experiences is the length of stay. A length of stay index for a tourist provides a clear and reliable
indicator for the level of consumption (and revenue) for tourist destinations. According to Rugg
(1973), the utility of traveling is obtained from staying at certain destinations for some period of
3
time, rather than the possessing or consuming destinations directly. Therefore, this index is often
used to measure the consumption level of tourists in a particular destination. As an indicator
reflecting the demands of individual tourists, the length of stay also shows the extent of tourism
demand at a particular destination. This index is of great value for destination management and
marketing. Generally, the longer tourists stay, the more services and goods they are expected to
consume. Therefore, visits of long duration are considered to be beneficial to local economies
through the multiplier effect of the tourism industry (Archer & Shea, 1975).
A substantial amount of literature has reported investigations of tourism demand using various
methods (Lim, 1997; Song & Li, 2008). However, most of these studies focused on macro tourism
demand with aggregated data. They regarded tourists in certain groups as homogeneous. Few
studies have concentrated on micro tourism demand analysis, which examined the tourism
demand function of the individual tourist. The socio demographic attributes of each tourist can be
taken into consideration by analyzing individual demand models, which are more reliable and
appropriate. Among studies on micro tourism demand analysis, many have selected destination
choice (Correia, Barros, & Silvestre, 2007; Hong, Kim, Jang, & Lee, 2006; Huybers, 2005) or
expenditures (GongSoog, Fan, Palmer, & Bhargava, 2005) as major focuses. In contrast, only a
few papers have investigated the issue of length of stay for tourism demand analysis.
To model the length of stay, several econometric methods have been adopted. In early papers,
the two stage square least square (2SLS) model was frequently used, considering the simultaneous
causality between tourists’ expenditure and length of stay (Silberman, 1985; Uysal, McDonald, &
O'Leary, 1988; Walsh & Davitt, 1983). More recently, to model the duration of stay, the Tobit
model (Fleischer & Pizam, 2002; J. Mak & Moncur, 1979), which focuses on the data generation
4
process for censored data, and survival analysis, which is powerful in dealing with censored
duration data (Barros, Correia, & Crouch, 2008; Gokovali, Bahar, & Kozak, 2007; Gomes de
Menezes, Moniz, & Cabral Vieira, 2008; MartínezGarcia & Raya, 2008), are applied. However,
in China, most tourists survey questionnaires follow the template from CNTA’s national domestic
tourists survey in which case the options of “duration of stay” are listed in ordinal scale. This
means that the options are dichotomous and ranked. Therefore, to appropriately use or apply this
data, an ordered logit model is more efficient and suitable, and has been chosen in this study.
Another contribution of this research to the current body of tourism marketing literature is the
introduction of generalized ordered logit model. Some authors have applied ordered logit model to
examine tourists’ preference to culture heritage sites (van Leeuwen & Nijkamp, 2010), consumer
meal preference (Myung, Feinstein, & McCool, 2008), visitors’ duration of theme park activities
(Kemperman, Borgers, Oppewal, & Timmermans, 2003), intended duration of students’
international travel (Chadee & Cutler, 1996), and tourists’ evaluation of different products (Cuccia
& Cellini, 2007). However, since the validity of ordered logit estimation relies heavily on the
parallel regression assumption of parameters, and this assumption is highly likely to violate in
empirical studies (Long & Fresse, 2003), more robust model specifications should be introduced
to alleviate this problem. The generalized ordered logit model, which relaxes this assumption, is a
natural alternative to the traditional ordered logit model (Williams, 2006).
The main purpose of this article is to identify the determinants of length of stay for tourists in
Yixing, a famous tourist destination in Eastern China, and to discuss the traveling behavior of
Chinese domestic tourists. Once the determinants are identified, more specific and efficient
marketing strategy can be applied accordingly. Since little research has focused on Chinese
5
domestic tourists’ behavior, especially research utilizing quantitative analysis, this paper will
provide one of the first attempts to do so, and to examine the differences between Chinese
domestic tourists and those in western countries. Furthermore, since a tourist’s duration of stay is
influenced by destination attributes, destinations and tourism enterprises could improve their
products and services according to the results of such research to influence tourists to extend their
stay, leading to improved revenue for local economies.
This paper is organized as follows: Section 2 discusses factors contributing to the length of stay.
Section 3 details the specifications of the model and describes the data set used in this study. In
Section 4, the empirical results of the model and interpretations are explained. Finally, concluding
remarks are presented in Section 5.
FACTORS CONTRIBUTING TO THE LENGTH OF STAY
To further discuss the past research on length of stay, several categories of determinants of
length of stay are respectively discussed.
Economic Variables According to the traditional demand theory, consumers’ income and the price
of a product determine the amount that an individual consumes. People with higher disposable
income are more likely to consume more commodities. Similarly, regarding the daily experience
at a particular destination as a single product, people with higher disposable income are inclined to
purchase more products, and therefore to stay longer in the destination (Fleischer & Pizam, 2002;
Gokovali, et al., 2007; J. Mak & Moncur, 1979; James Mak, Moncur, & Yonamine, 1977). On the
other hand, the price of the product has a negative effect on the duration of stay for tourists
6
(Alegre & Pou, 2006, 2007; Gomes de Menezes, et al., 2008; J. Mak & Moncur, 1979; James Mak,
et al., 1977; Silberman, 1985; Walsh & Davitt, 1983).
Distance The distance from the tourist’s residence to the destination is another important factor
influencing the length of stay. In order to obtain higher utility from a trip, a rational tourist is
expected to balance the proportion between the fixed cost and the varied cost (Smith, 1995).
Therefore, the tourist traveling from a more distant place of residence with a higher fixed cost will
increase the varied cost in destinations for a longer duration, to optimize the proportion between
the two components (Thunberg & Crotts, 1994; Walsh & Davitt, 1983).
Socialdemographic Variables Several socialdemographic variables are associated with the length
of stay. Tourists of different ages have different propensities on the length of their stay, and the
relationship between the length of stay and age is nonlinear (Alegre & Pou, 2006; Fleischer &
Pizam, 2002; Gokovali, et al., 2007; James Mak, et al., 1977). For example, Fleischer and Pizam
(2002) found that for Israel seniors between the ages of 55 and 65, an increase of leisure time and
household income resulted in an increase in the number of vacation days. For those older than 65,
the declination of income and the deterioration of health results in a decrease in the number of
vacation days. Furthermore, family life cycle influences tourists’ duration choice (Oppermann,
1995). Lawson (1991) found that young singles (younger than twentyfive) stay the longest time,
while young couples (no children) and people in the "Empty Nest I" stage (still working, no
children) have short duration.
7
Past Traveling Experience Past traveling experience of the tourist plays an indispensable role in
determining the duration of stay. First, familiarity gained from past traveling experience influences
the decision of duration. Repeat visitors always stay longer in a single destination than firsttime
visitors (Oppermann, 1997; Uysal & McDonald, 1989). Wang (2004) stated that, due to a higher
level of awareness and familiarity of the destination, repeat visitors tend to have different
itineraries or possible emotional attachment to a place, resulting in longer duration. Furthermore,
annual traveling frequency has also been reported as a determinant of length of stay (Fleischer &
Pizam, 2002). As the experience of tourism increases, tourists are more inclined to stay longer in a
destination (Gokovali, et al., 2007).
Trip Characteristics Different traveling motivations result in distinct duration. For different
purposes, tourists participate in different activities and make different itineraries for trips, which
require different duration (Andreu, Kozak, Avci, & Cifter, 2005; Kim & Prideaux, 2005; Seaton &
Palmer, 1997). For example, tourists traveling for VFR purposes stay longer in a destination than
those traveling for other purposes (Hsu & Kang, 2007; Sung, Morrison, Hong, & O’Leary, 2001),
while the tourist on a sightseeing trip stays for a shorter period of time. Furthermore, it has been
found that tourists on an organized package tour have a greater possibility of staying longer than
individual tourists (Alegre & Pou, 2007). Other vacation characteristics that influence the length
of stay include traveling party size (Alegre & Pou, 2006), accommodation type (Gokovali, et al.,
2007), and type of board (Alegre & Pou, 2006, 2007).
Destination Related Variables Gokovali, Bahar and Kozak (2007) emphasized the importance of
8
satisfaction during the stay in a destination. It was argued that, as long as tourists make the
decision to stay longer, they become more aware of the facilities and services at their destination.
Also, the perceived attractiveness and image of the destination affect the duration of stay
(Gokovali, et al., 2007). Moreover, through a study of travelers in Colorado, Tierney (1993) found
that a welcome center contributes to a longer length of stay. Other destination attributes, such as
annual rainfall (J. Mak & Moncur, 1979) and crowding (Uysal, et al., 1988), also exert influences
on the length of stay for tourists in particular destinations.
METHODOLOGY
Specified Model and Data
After reviewing previous research, this study proposes an empirical model for analyzing the
potential factors which may influence a tourists’ duration of stay. The model is as follows:
), ,,,,,,( 2assessmentvisitpastmotivationtiontransportaorganizeddistanceageageflength
In this model, the dependent variable is the length of stay for each tourist. This variable is coded
as: 1 for a oneday stay, 2 for a twoday stay, 3 for a threeday stay, and 4 for a fourday or longer
stay. The selection of explanatory variables is based on the relevant findings in past research.
Three categories of variables are considered. The first category includes variables representing
individual attributes, such as age. It is assumed that age exerts a nonlinear influence on the
duration of tourists’ stays, and the squared term of age is included in the model. This means that
below a certain age, the duration of stay increases as age increases. However, above a certain age,
the duration declines as age increases. The second category of variables is comprised of trip
characteristic variables, such as distance, organized, transportation, motivation, and past visit. The
9
third category includes destination related variables. In the model assessment is the perceived
assessment of a particular type of service in the destination, such as the assessment of
accommodation, transportation, tour guides, entertainment, and shopping. Table 1 summarizes
variables used in this study and their reference categories.
(Insert Table 1 here)
This study chose Yixing, a city in Jiangsu Province in Eastern China (Figure 1), to conduct the
empirical work. There are several reasons. First, as the members of Yixing tourism master planning
group, it is convenient for us to conduct the tourists’ survey. Second, Yixing is located in the center
of the Yangtze River Delta, which is the most developed areas in China and a great domestic
tourism market. Therefore, through the study of Yixing, we can know about the traveling behavior
of tourists from this market. Third, Yixing is a representative tourist destination in China with
diverse tourist attractions which attracts distinctively heterogeneous tourists. Famous tourist
attractions in Yixing include ‘Shanjuan Cavern’, a large limestone cavern; ‘Sea of Bamboo’, which
is renowned for natural sights, and many historic sites for historical events and ancient kiln sites.
According to the statistics from Yixing Tourism Bureau, the motivation of tourists to Yixing covers
a large spectrum, and its source market covers a broad geographical area. The heterogeneity of
tourists provides an ideal sample to estimate the empirical model.
(Insert Figure 1 here)
10
The research data was gathered from tourist survey in major tourist destinations in Yixing within
the period of JulyAugust 2004. The final sample included the valid responses from 417 visitors
who finished the questionnaire. The traveling distance of tourists is obtained by measuring the
road distance from Yixing to tourist’s residence reported in the questionnaire through ArcGIS@
software. Descriptive statistics of variables are provided in Tables 2A and 2B. In Table 2A, the
average age of tourists is 35.57 and the average traveling distance is 432.37 kilometers. All
destination attribute variables for the assessment of service quality are lower than 3
(corresponding to the level of “good”), indicating that tourist satisfaction is not very high. As
shown in Table 2B, more than 60% of tourists to Yixing are oneday travelers, while only about
7% of tourists stay in Yixing more than four days. They come to Yixing mainly by coach/bus or in
their own vehicles (selfdriving). Moreover, most tourists, accounting for nearly 80% of total
tourists, travel for sight seeing or vacation. Also, more than 80% of tourists have traveled to Yixing
less than two times.
(Insert Table 2A here)
(Insert Table 2B here)
Econometric Method
As previously mentioned, the ordered logit model and its extension, the generalized ordered
logit model, will be estimated in this research. This model is commonly presented as a latent
variable model. Deﬁning y∗ as a latent variable ranging from −∞ to +∞, the structural model is
ii
y
i
X
*
11
where i is the observation (individual tourist) and ε is a random error (Long & Fresse, 2003). The
measurement model for divide y
∗
into J ordinal categories:
myi
if
mm y
*
1
for
1m
to
J
where the cutpoints τ 1 through τJ−1 are estimated. We assume τ0 = −∞ and τJ = +∞.
In this paper, responses in the model are: 1 for a oneday stay, 2 for a twoday stay, 3 for a
threeday stay, and 4 for a fourday or longer stay. Hence, J = 4 in this model. The continuous
latent variable can be thought of as the propensity to stay longer in the destination. To include
independent variables, we can rewrite the probability of duration as a function of explanatory
variables:
)()()Pr( 1
mm FFmy
where F(.) is a logit function. To further explain the results from ordered logit model, the concept
of odds is introduced to interpret the coefficient:
)exp(
)Pr( )Pr(
)(

mmm my my
odds
The generalized ordered logit model, as an extension to the traditional ordered logit model, allows
β
to differ for each of the J1 comparisons (Williams, 2006). That is,
)()()(Pr* 11 mmmm FFmy
In this case, the odds is no longer a constant among different outcomes in the dependent variable,
and it becomes:
)exp(
)(Pr* )(Pr*
)(* mmmm my my
odds
which varied between different outcomes of dependent variables.
12
EMPIRICAL RESULTS
First, ordered logit models with different specifications are first estimated using Stata@ 12.0
package. Table 3 summarizes results from specified models which include different assessment
variables. The result shows that nearly all variables included in the models have significant effects
on the length of stay and their signs are generally in line with findings reported in previous studies.
In terms of their explanatory power, both equations are satisfactory when the likelihood ratio (LR)
test and log likelihood diagnostic statistics are similarly acceptable. Moreover, the estimated
coefficients in each model vary little, demonstrating the ideal robustness of each model.
The results in Table 3 suggest that distance is a significant determinant of duration. Tourists
from source markets more distant from the destination tend to stay longer. With respect to age, as
hypothesized, both age and the squared term of age (age_squared) are significant. It indicates the
nonlinear relationship between age and duration: the length of stay would not keep on rising as
the age of the tourist increases; the possibility for long duration would decline as the age increases
above a certain threshold value. For multiple response questions in the model, after setting one
variable as the reference, others are coded into dummy variables. In this study, the Wald test is
applied to test the overall significance of such multiple response questions as transportation,
motivation, and past visit. The result shows that they are all significant at a 0.05 level. Therefore,
we can conclude that transportation, motivation, and past visit are significantly associated with
tourists’ duration of stay. Given the estimated magnitude of the coefficient of each dummy
variable, we can infer which categories of tourists stay significantly longer than others. With
respect to transportation, tourists traveling by train and airplane have greater lengths of stay in
Yixing than those traveling by coach/bus and selfdriving. Also, tourists with different motivations
13
have different lengths of stay. Visitors for research purpose stay longest while those with
sightseeing purposes stay shortest, and visitors with VFR and study purposes exhibit greater
lengths of stay than those visiting for vacation. Moreover, with respect to past visit, tourists who
have been to Yixing more than five times stay significantly longer than firsttime visitors.
Furthermore, in Table 3, Models 1 to 5 are compared to investigate which assessment variable is
more important for determining tourists’ duration of stay. It is suggested that assessment of
accommodation is the only factor that exerts influence on the length of stay based on the
significance of the coefficient and the goodnessoffit indices, such as log likelihood and LR value.
Therefore, it is suggested that accommodation assessment is a statistically significant determinant
of length of stay.
(Insert Table 3 here)
Attention is now turned to the duration model for different tourists. In Table 4, Models 6 and 7
estimate the ordered logit model for organized tourists and individual tourists respectively; while
Models 8, 9, and 10 estimate the model for tourists aged less than 30 (Age group 1), between 31
and 40 (Age group 2), and more than 41 (Age group 3). In each model, the only variable included
is the “assessment of accommodation” so as to avoid the problem of multicollinearity which
occurred when all assessment variables were employed. Looking first at Models 6 and 7,
interestingly, it shows that the age and accommodation assessment are not determinants of the
length of stay for organized tourists. This may reflect the fact that organized tourists have much
less flexible schedules for traveling, and their itineraries are determined by others such as the
14
travel agency or the group that organized the trip. Besides, individual tourists stay significantly
longer for VFR purposes than for sightseeing purposes; while organized tourists do not. On the
other hand, organized tourists on vacation stay longer than sightseeing organized tourists, while
their counterparts, traveling individually, do not.
In regard to models for different age groups, the group of motivation variables is jointly
significant in Models 8 and 10, while not in Model 9. For those in Age Groups 1 and 3, VFR
tourists stay significantly longer than sightseeing tourists, while it is not true in Age Group 2.
Furthermore, past visit to Yixing would significantly increase tourists’ probabilities to stay longer
for tourists in Age Groups 1 and 3. However, in Age Group 2, the difference of the length of stay
only exists between firsttime visitors and repeated visitors, but not between frequently visited
tourists (3 times and above) and less frequently visited tourists (1 time to 2 times). Moreover, the
result suggests that there is not significant difference in the length of stay for organized and
individual tourists in Age Group 3. This may be explained by the fact that tourists in this age
group are more likely to undertake longer organized visit either by purchasing package tour with
longer duration or by attending tours organized by affiliations as incentive tours which is
characterized by longer durations. As a result, tourists in organized tours may not stay
significantly shorter than individual ones. Also, the accommodation assessment is not associated
with the duration for tourists in this age group. This can be contributed by the fact that aged
domestic tourists in China are not demanding in regard to accommodation service and facility due
to their life experience in early days with poor living conditions, especially in the period before
the openingup.
15
(Insert Table 4 here)
However, as indicated in Table 3, the approximate LR test for parallel regression assumption
suggests that the parallel regression assumption is violated in previous models. The Wald test by
Brant (1990) is used to test the parallel regression assumption for each variable individually. The
result of the test argues that the largest violations are for organized visitation and assessment of
accommodation, which implies that these two variables are misspecified in the traditional ordered
logit model. Therefore, the generalized ordered logit model, not imposing the constraints of
parallel regressions for these two variables, is adopted to further explore potential determinants of
length of stay in Model 11. In Model 11, age, square of age, transportation variables, motivation
variables, and past visitation variables are constrained with the parallel assumption, which means
the coefficients of these variables are the same for each outcome category (each category of length
of stay). For the unconstrained variables, organized visitation and assessment of accommodation,
the marginal effects for each outcome category are not the same, and coefficients are estimated for
each category.
The estimation result of Model 11 is presented in Table 5. The estimated coefficients of most
constrained variables do not differ too much from the result in Table 3. The greatest difference is
the estimated coefficients of unconstrained variables. Organized visitation is only significant for
the outcome categories "1 day vs. 2 days and more" and "12 day vs. 3 days and more." Therefore,
it is suggested that organized visitation only influences the length of stay for tourists with short
duration. In the same way, judging from the estimated coefficient, longer duration tends to be
more sensitive to the assessment of accommodation. This reveals the fact that tourists’ assessment
16
of accommodation is a vital factor for stimulating longer duration of stay.
(Insert Table 5 here)
To make the interpretation more straightforward, the estimation result of each coefficient is
explained based on the concept of odds. Based on the result from Model 11, for a unit increase in
distance, the odds of a shorter length of stay outcome compared with a longer length outcome are
changed by the factor exp(0.0004)=0.9996, holding all other variables constant. For nominal
variables, the change of transportation from coach/bus to selfdriven would contribute to a
exp(((2.342)(1.707)))=1.887 change of odds of a shorter length of stay outcome compared with
a longer length, while a change from train to selfdriven would cause a exp((0.3530))=1.423
change. Likewise, a change of motivation from VFR to vacation will cause an
exp((0.6881.687))=2.716 change of those odds, and a change of past visitation from firsttime
visit to 3 time or 4 times will contribute to an exp(((1.309)( 1.872)))=0.569 change.
To further elaborate the estimation, graphing predicted probabilities for each outcome is
employed. Figure 2 plots the predicated probabilities of each outcome of length of stay with the
distance change from Model 11. It emphasizes the positive effect of distance on length of stay,
showing that the probability of a 1 day stay decreases dramatically as traveling distance increases,
while the probabilities of longer durations increase respectively. Since parallel regression
assumption is imposed on the coefficient of distance, the probability line in the graph is straight,
suggesting that the marginal effects of distance on each duration category is constant.
17
(Insert Figure 2 here)
The graphic interpretation is also useful for explaining the nonlinear effects in the model with
age and age_squared. Figure 3 demonstrates the nonlinear effect of age on the length of stay
based on the result from Model 11. It shows that there is a turning point in the probability curve at
the age of 45. Therefore, it is suggested that when age is less than 45, the probability of long
duration decreases while the probability of a one day additional stay increases with age. However,
the situation is inverse when the age of the tourist is more than 45. This result is consistent with
findings from Alegre and Pou (2006), Fleischer and Pizam (2002), Gokovali, Bahar and Kozak
(2007).
(Insert Figure 3 here)
Figure 4 depicts the nonparallel marginal effect of assessment of accommodation on duration.
Due to the relaxation of parallel constraints, the probability line in the graph is no longer a strict
strait line (in contrast to Figure 2); instead, there are several minor turning points on the line. In
this figure, it is suggested that when the assessment of accommodation changes from ‘good’ to
‘fairly good,’ the probability of a ‘4 days or more’ stay increases significantly. As previously
discussed, longer duration is considered to be more sensitive to the assessment of accommodation.
(Insert Figure 4 here)
18
An important advantage of the (generalized) ordered logit models is that they can be utilized for
simulation after the estimation. By holding other variables at average value, the effect of
explaining a variable on the probability can be drawn. Table 6 shows several simulation results in
diverse scenarios. For example, if a tourist’s traveling distance changes from 100km to 300km,
his/her probability of a 1 day stay would decrease 1.770%, while the probability of a 2 days stay, 3
days stay, and 4 days or more stay would increase 1.070%, 0.480% and 0.220%, respectively,
holding other variables constant. In the same way, as shown in Table 6, if a tourist changes the
motivation from sightseeing to VFR, the probability of a 1 day stay would decline by 39.15%,
and at the same time the probability of 2 days stay, 3 days stay and 4 days or more stay would
increase 9.410%, 3.700% and 1.600%.
(Insert Table 6 here)
CONCLUSION
The purpose of this study is to examine the effect of various factors on a tourist’s length of stay.
The ordered logit model is utilized to estimate potential determinants. The findings of this study
show traveling distance, age (with the squared term of age), organized visits, transportation,
motivation, past visits, and assessment of accommodation to be influential factors. Furthermore,
some differences in duration models are found between organized tourists and individual tourists.
Organized tourists are not influenced by age and assessment of accommodation, while individual
19
tourists are. Moreover, tourists with different ages determine their length of stay differently:
organized visitation and assessment of accommodation are not factors affecting the duration for
tourists above 41 years of age. To elaborate the estimation result, the study analyzes the
relationship between age and length of stay by using the probability the model calculated, and the
nonlinear effect is demonstrated.
Based on the model, some marketing suggestions and implications can be proposed. To begin
with, extra attention should be paid to individual tourists to attract them to stay longer in a given
destination. Also, to attract VFR visitors, and to make them stay longer in Yixing, more family
activities and products are recommended. In addition, since repeat visitors are very important,
extra effort should be allocated on how to improve service quality and develop tourists’
destination loyalty, especially for older tourists. Finally, as indicated by the model, the assessment
of accommodation is crucial for tourists’ length of stay, especially for stays of longer duration. It is
necessary for local government and industry to improve both the hard and soft infrastructure of
accommodation facilities.
Finally, the limitation of this study should be stated. Due to the unavailability of data, variables
such as income and daily expenditures could not be obtained and included in the estimated model.
Moreover, since this research only focuses on a single destination, further studies should be
undertaken to investigate how the various attributes in different destinations would influence the
length of stay for tourists. Understanding these factors allows those responsible for an area’s
tourism to make a destination more inviting to tourists, and this could bring about a higher rate of
trade and improved revenue for local businesses located in close proximity to the destination.
20
REFERENCES
Alegre, J., & Pou, L. (2006). The length of stay in the demand for tourism. Tourism Management, 27(6),
13431355.
Alegre, J., & Pou, L. (2007). Microeconomic determinants of the duration of stay of tourists. In Á.
Matias, P. Nijkamp & P. Neto (Eds.), Advances in Modern Tourism Research (pp. 181206).
Heidelberg: PhysicaVerlag
Andreu, L., Kozak, M., Avci, N., & Cifter, N. (2005). Market segmentation by motivations to travel:
British tourists visiting Turkey. Journal of Travel & Tourism Marketing, 19(1), 114.
Archer, B. H., & Shea, S. (1975). Length of stay problems in tourist research. Journal of Travel
Research, 13(3), 810.
Barros, C. P., Correia, A., & Crouch, G. (2008). Determinants of the length of stay in Latin American
tourism destinations. Tourism Analysis, 13(4), 329340.
Chadee, D. D., & Cutler, J. (1996). Insights into international travel by students. Journal of Travel
Research, 35(2), 7580.
Correia, A., Barros, C. P., & Silvestre, A. L. (2007). Golf tourism repeat choice behaviour in the
Algarve: a mixed logit approach. Tourism Economics, 13(1), 111127.
Cuccia, T., & Cellini, R. (2007). Is cultural heritage really important for tourists? A contingent rating
study. Applied Economics, 39(2), 261271.
Fleischer, A., & Pizam, A. (2002). Tourism constraints among Israeli Seniors. Annals of Tourism
Research, 29(1), 106123.
Gokovali, U., Bahar, O., & Kozak, M. (2007). Determinants of length of stay: a practical use of
survival analysis. Tourism Management, 28(3), 736746.
21
Gomes de Menezes, A., Moniz, A., & Cabral Vieira, J. (2008). The determinants of length of stay of
tourists in the Azores Tourism Economics, 14(1), 205222.
GongSoog, H., Fan, J. X., Palmer, L., & Bhargava, V. (2005). Leisure Travel Expenditure Patterns by
Family Life cycle Stages. Journal of Travel & Tourism Marketing, 18(2), 1530.
Hong, S.k., Kim, J.h., Jang, H., & Lee, S. (2006). The roles of categorization, affective image and
constraints on destination choice: An application of the NMNL model. Tourism Management,
27(5), 750761.
Hsu, C. H. C., & Kang, S. K. (2007). CHAIDbased segmentation: International visitors' trip
characteristics and perceptions. Journal of Travel Research, 46(2), 207216.
Huybers, T. (2005). Destination choice modelling: what's in a name? Tourism Economics, 11(3),
329350.
Kemperman, A., Borgers, A., Oppewal, H., & Timmermans, H. (2003). Predicting the duration of
theme park visitors' activities: an ordered logit model using conjoint choice data. Journal of
Travel Research, 41(4), 375384.
Kim, S. S., & Prideaux, B. (2005). Marketing implications arising from a comparative study of
international pleasure tourist motivations and other travelrelated characteristics of visitors to
Korea. Tourism Management, 26(3), 347357.
Lawson, R. (1991). Patterns of tourist expenditure and types of vacation across the family life cycle.
Journal of Travel Research, 29(4), 1218.
Lim, C. (1997). Review of international tourism demand models. Annals of Tourism Research, 24(4),
835849.
Long, S., & Fresse, J. (2003). Regression Models for Categorical Dependent Variables Using STATA
22
(Revised Edition ed.). Texas, USA: STATA Press.
Mak, J., & Moncur, J. (1979). The choice of journey destinations and length of stay: a micro analysis.
The Review of Regional Studies, 10(2), 3848.
Mak, J., Moncur, J., & Yonamine, D. (1977). Determinants of visitor expenditures and visitor lengths
of stay: a crosssection analysis of U.S. visitors to Hawaii. Journal of Travel Research, 15(3),
58.
MartínezGarcia, E., & Raya, J. M. (2008). Length of stay for lowcost tourism. Tourism Management,
29(6), 10641075.
Myung, E., Feinstein, A. H., & McCool, A. C. (2008). Using a discrete choice model to identify
consumer meal preferences within a prix fixe menu. Journal of Hospitality & Tourism
Research, 32(4), 491504.
Oppermann, M. (1995). Travel life cycle. Annals of Tourism Research, 22(3), 535552.
Oppermann, M. (1997). Firsttime and repeat visitors to New Zealand. Tourism Management, 18(3),
177181.
Rugg, D. (1973). The choice of journey destination: a theoretical and empirical analysis. The Review of
Economics and Statistics, 55(1), 6472.
Seaton, A., & Palmer, C. (1997). Understanding VFR tourism behaviour: the first five years of the
United Kingdom tourism survey. Tourism Management, 18(6), 345355.
Silberman, J. (1985). A demand function for length of stay: the evidence from Virginia beach. Journal
of Travel Research, Spring, 1623.
Smith, S. L. J. (1995). Tourism Analysis: A Handbook (2nd ed.). Harlow: Longman.
Song, H., & Li, G. (2008). Tourism demand modelling and forecasting—a review of recent research
23
Tourism Management, 29(2), 203220.
Sung, H. H., Morrison, A. M., Hong, G.S., & O’Leary, J. T. (2001). The effects of household and trip
characteristics on trip types: a consumer behavioural approach for segmenting the US
domestic leisure travel market. Journal of Hospitality and Tourism Research, 25(1), 4668.
Thunberg, E. M., & Crotts, J. C. (1994). Factors affecting travelers' overnight stay behavior. Journal of
Travel & Tourism Marketing, 3(1), 117.
Tierney, P. T. (1993). The influence of state traveler information centers on tourist length of stay and
expenditures. Journal of Travel Research, 31(3), 2832.
Uysal, M., & McDonald, C. D. (1989). Visitor segmentation by trip index. Journal of Travel Research,
27(3), 3842.
Uysal, M., McDonald, C. D., & O'Leary, J. T. (1988). Length of stay: a macro analysis for
crosscountry skiing trips. Journal of Travel Research, 26(3), 2931.
van Leeuwen, E., & Nijkamp, P. (2010). A microsimulation model for eservices in cultural heritage
tourism Tourism Economics, 16(2), 361384.
Walsh, R. G., & Davitt, G. J. (1983). A demand function for length of stay on ski trips to Aspen. Journal
of Travel Research, 21(4), 2329.
Wang, D. (2004). Tourist behaviour and repeat visitation to Hong Kong. Tourism Geographies, 6(1),
99118.
Williams, R. (2006). Generalized ordered logit/partial proportional odds models for ordinal dependent
variables The Stata Journal, 6(1), 58–82.
24
Table 1
Summary of Explanatory Variables
Name
Explanation
1. Individual attributes
age
Age of the reference person (in years)
age_squared
Squared term of age
income
Monthly income (in 1,000 RMB)
frequency
Number of domestic tour in last year (1 = 12 tours (reference),
2 = 35 tours, 3 = 6 tours and more)
2. Trip characteristics
distance
Distance between Yixing and the residence of tourist (in 100
kilometers)
partner
1 if alone (reference), 2 if with friends and relatives, 3 if with
colleagues, 4 if organized by travel agency
transportation
1 if selfdriving (reference), 2 if coach/bus, 3 if airplane, 4. if
train, 5 if others
pastvisit
1 if no previous visit (reference), 2 if one to two times, 3 if
three to four times, 4 if five times and above
motivation1
1 if sightseeing for motivation, 0 otherwise
motivation2
1 if vacation for motivation, 0 otherwise
motivation3
1 if VFR for motivation, 0 otherwise
motivation4
1 if business for motivation, 0 otherwise
3. Destination related variables
assessment of accommodation
1 = bad, 2 = neutral, 3 = good, 4 = fairly good,
assessment of transportation
1 = bad, 2 = neutral, 3 = good, 4 = fairly good,
assessment of tour guides
1 = bad, 2 = neutral, 3 = good, 4 = fairly good,
assessment of entertainment
1 = bad, 2 = neutral, 3 = good, 4 = fairly good,
assessment of shopping
1 = bad, 2 = neutral, 3 = good, 4 = fairly good,
25
Table 2A
Descriptive Statistics of Continuous Variables
Variable
Mean
Std. Dev.
Min
Max
distance
4.25
6.09
0.53
55.39
age
33.39
11.65
9
65
income
29.97
45.55
4.5
700
assessment of accommodation
2.65
0.78
1
4
assessment of transportation
2.67
0.79
1
4
assessment of tour guides
2.77
0.83
1
4
assessment of entertainment
2.49
0.71
1
4
assessment of shopping
2.36
0.83
1
4
Table 2B
Descriptive Statistics of Discrete Variables
Variable
Category
Frequency
Percentage
Cumulated
Percentage
LOS
1 day
243
55.23
55.23
2 days
115
26.14
81.36
3 days
45
10.23
91.59
4 days
8
1.82
93.41
5 days and more
29
6.59
100
frequency
12 tours
226
51.36
51.36
35 tours
165
37.50
88.86
6 and more tours
49
11.14
100
partner
alone
72
16.36
16.36
with friends and re
latives
258
58.64
75
with colleagues
97
22.05
97.05
organized by travel
agency
13
2.95
100
transportation
selfdriving
164
37.27
37.27
coach/bus
226
51.36
88.64
airplane
7
1.59
90.23
train
31
7.05
97.27
others
12
2.73
100
past visit
No previous visit
227
51.59
51.59
1 time to 2 times
142
32.27
83.86
3 to 4 times
35
7.95
91.82
5 times and above
36
8.18
100
motivation1
sight seeing
165
37.50
motivation2
vacation
175
39.77
motivation3
VFR
39
8.86
motivation4
business
28
6.36
26
TABLE 3
Model
Data generation process
Conditional expectation
Likelihood function for right censored data
Linear
Regression (with
logarithm
transformation)
2
ln
(0, )
i i i
y
N
x
(ln  )
i i i
Ey
xx
1
2
2
2
11
exp (ln ) 1
2
2
dd
i
ii c
y
x
x
Poisson
Regression
Pr(  ) !
exp( )
im
i
ii
ii
e
ym m
x
x
ln (  )
i i i
Ey
xx
1
exp( )
exp( )
exp( )
exp( )
!
1
!1
i
ii
i
i
d
j
d
yjc
i
ej
ee
ye
x
x
x
x
Weibull model
(AFT metric)
0
ln ln
Weibull( , )
i i i
y
p
x
1
(ln  )
i i i
Ey p
xx
1
exp( ) ( exp( ) ) exp [ exp( ) ]
d
pp
i i i i i
p y y
x x x
Lognormal
model (AFT
metric)
0
ln ln
Lognormal( , )
i i i
y
x
(ln  )
i i i
Ey
xx
1
ln( exp( ) ) ln( exp( ) )
1()
dd
i i i i
i
yy
y
xx
Loglogistic
model (AFT
metric)
0
ln ln
Loglogistic( , )
i i i
y
x
(ln  )
i i i
Ey
xx
11
11
1
exp( ) [ exp( ) ] 1
1 exp( ) 1 exp( )
d
i i i
i i i i
y
yy
xx
xx
(NOTE: D )
27
28
Table 3
Estimation Results of Ordered Logit Model with Different Specifications
Variable
Model 1
Model 2
Model 3
Model 4
Model 5
dist
0.0104**
0.0127***
0.0178**
0.0155**
0.00890**
0.0104***
0.0111**
(0.00405)
(0.00420)
(0.00787)
(0.00611)
(0.00449)
(0.00405)
(0.00445)
age
0.0422***
0.0562***
0.109***
0.0799***
0.0556***
0.0422***
0.0399***
(0.0128)
(0.0149)
(0.0264)
(0.0203)
(0.0159)
(0.0128)
(0.0144)
age_sq
0.000494***
0.000641***
0.00127***
0.000931***
0.000663***
0.000494***
0.000471**
(0.000177)
(0.000203)
(0.000362)
(0.000279)
(0.000219)
(0.000177)
(0.000200)
income
0.000888**
0.00131**
0.00260*
0.00168
0.000815
0.000888**
0.000844**
(0.000390)
(0.000564)
(0.00156)
(0.00108)
(0.000673)
(0.000390)
(0.000363)
transport_3
0.133***
0.152**
0.306**
0.235**
0.187***
0.133***
0.121**
(0.0509)
(0.0618)
(0.131)
(0.0956)
(0.0720)
(0.0509)
(0.0563)
transport_4
0.656***
0.659***
1.318***
0.944***
0.648***
0.656***
0.713***
(0.180)
(0.165)
(0.286)
(0.224)
(0.168)
(0.180)
(0.191)
transport_5
0.642***
0.645***
1.107***
0.894***
0.692***
0.642***
0.678***
(0.101)
(0.105)
(0.183)
(0.136)
(0.127)
(0.101)
(0.104)
transport_6
0.274**
0.227*
0.504**
0.430**
0.233*
0.274**
0.361**
(0.133)
(0.135)
(0.234)
(0.178)
(0.140)
(0.133)
(0.146)
partner_n_2
0.228***
0.314***
0.688***
0.491***
0.308***
0.228***
0.209***
(0.0666)
(0.0824)
(0.189)
(0.135)
(0.0877)
(0.0666)
(0.0704)
partner_n_3
0.142*
0.267***
0.628***
0.412**
0.190*
0.142*
0.129*
(0.0745)
(0.0980)
(0.226)
(0.160)
(0.0994)
(0.0745)
(0.0765)
partner_n_4
0.0231
0.0781
0.295
0.187
0.123
0.0231
0.00215
(0.144)
(0.176)
(0.471)
(0.328)
(0.205)
(0.144)
(0.145)
pastvisit_2
0.0906*
0.112*
0.199
0.159
0.0668
0.0906*
0.0960*
(0.0529)
(0.0665)
(0.141)
(0.104)
(0.0666)
(0.0530)
(0.0551)
pastvisit_3
0.346***
0.329***
0.562**
0.477***
0.430***
0.346***
0.361***
(0.0945)
(0.113)
(0.225)
(0.163)
(0.115)
(0.0945)
(0.106)
pastvisit_4
0.457***
0.549***
1.005***
0.770***
0.500***
0.457***
0.500***
(0.104)
(0.112)
(0.179)
(0.142)
(0.116)
(0.104)
(0.126)
(*** indicates significant at p<0.01, ** indicates significant at p<0.05, * indicates significant at p<0.1)
29
Table 4
Estimation Results of Ordered Logit Model with Different Samples
Variable
Model 61
(Organized)
Model 7
(Individual)
Model 8
(Age Group 1)
Model 92
(Age Group 2)
Model 10
(Age Group 3)
distance
0.000
0.000
0.000
0.001**
0.000
age
0.142
0.111*
age_squared
0.002
0.001*
organized
0.948**
1.293**
0.643
transportation
(df)
8.09(2)**
34.90(3)***
20.18(3)***
7.04(2)**
8.14(3)**
coach/bus
2.559***
1.359***
2.480***
1.258
0.986
selfdriving
2.218**
2.407***
2.505***
2.476**
1.658**
airplane
1.099
－1.522
1.376
train＊
0.000
0.000
0.000
0.000
0.000
motivation
(df)
30.60(5)***
32.14(5)***
33.00(5)***
8.24(5)
10.23(5)*
sightseeing＊
0.000
0.000
0.000
0.000
0.000
vacation
1.511***
0.365
0.107
1.053**
1.098**
VFR
0.517
2.167***
1.897***
0.765
2.008***
business
0.919
1.068*
0.946
0.221
0.602
research
4.727***
0.893
5.073***
3.974**
1.018
study
1.358*
4.753***
1.641**
1.308
1.207
past visit
(df)
13.35(3)***
9.70(3)**
12.36(3)***
4.89(3)
15.10(3)***
no visit
2.087***
1.473***
1.735***
2.091**
2.210***
1 to 2 times
1.668***
0.657
1.331*
1.005
1.884***
3 to 4 times
0.022
0.963
0.065
1.296
0.194
5 times and
above＊
0.000
0.000
0.000
0.000
0.000
accommodation
assessment
0.412
0.377*
0.698***
0.937***
－0.276
constant1
3.614
3.201***
1.388
0.958
2.692**
constant2
1.732
1.549
0.153
2.411
0.382
constant3
1.118
0.259
1.077
5.277***
1.153
log likelihood
114.226
232.985
145.539
86.728
111.196
number of obs.
157
259
160
126
130
LR (df)
73.18(14)***
91.82(15)***
75.06(14)***
28.67(13)***
45.92(14)***
(*** indicates significant at p<0.01, ** indicates significant at p<0.05, * indicates significant at p<0.1, and
＊
indicates the reference category in a group of dummy variables, whose coefficient is fixed to be 0)
1
Only one case in this sample chooses airplane to destination. So this case and the variable of airplane are
excluded from Model 6.
2
Only one case in this sample chooses airplane to destination. So this case and the variable of airplane are
excluded from Model 9.
30
Table 5
Estimation Results of Generalized Ordered Logit Model
Model 11
1 day vs.2 days
and more
12 day vs. 3
days and more
13 days vs. 4
days and more
distance
0.0004*
age
0.137***
age_squre
0.002***
organized
0.778***
0.876**
0.0009
transportation (df)
33.17(3)***
coach/bus
1.707***
selfdriven
2.342***
airplane
0.353
train＊
0.000
motivation (df)
45.13(5)***
sightseeing＊
0.000
vacation
0.688***
VFR
1.687***
business
1.016**
research
3.449***
study
1.596***
past visit (df)
27.71(3)***
firsttime visit
1.872***
1 to 2 times
0.446
3 to 4 times
1.309***
5 times and above＊
0.000
assessment of
accommodation
0.273*
0.569***
0.887***
constant
2.689**
0.143
0.597
log likelihood
365.229
number of obs.
417
LR (df)
111.02(20)***
Pseudo Rsqure
0.163
(*** indicates significant at p<0.01, ** indicates significant at p<0.05, * indicates significant at p<0.1, and
＊
indicates the reference category in a group of dummy variables, whose coefficient is fixed to be 0)
31
Table 6
Simulation Results from the Model
Scenario
Probability changes
1 day
2 days
3 days
4 days and
more
Distance: 100km to 300km
1.770%
1.070%
0.480%
0.220%
Age: 30 to 40
6.590%
4.070%
1.750%
0.770%
Assessment of accommodation:
neutral to good
6.390%
1.060%
2.790%
2.550%
Individual tourist to organized
one
12.750%
8.950%
3.800%
0.000%
Transportation: coach/bus to
selfdrive
14.160%
8.770%
3.730%
1.650%
Motivation: sighseeing to
vacation
14.710%
9.410%
3.700%
1.600%
Motivation: sightseeing to VFR
39.150%
19.420%
12.990%
6.740%
Past visitation: firsttime to 1
time and 2 times
12.660%
7.820%
3.360%
1.490%
32
Figure 1
Location Map of Yixing
33
Figure 2
Probability Change of Length of Stay with Distance
34
Figure 3
Probability Change of Length of Stay with Age
35
Figure 4
Probability Change of Length of Stay with Assessment of Accommodation