Content uploaded by Yang Yang
Author content
All content in this area was uploaded by Yang Yang on Oct 07, 2017
Content may be subject to copyright.
Domestic tourism demand of urban and rural residents in China:
Does relative income matter?
Yang Yang a
a Department of Geography, University of Florida, Gainesville, FL, 32611, USA
(Email: yang.yang@ufl.edu)
Ze-Hua Liu b
b Department of Land Resources and Tourism Sciences, Nanjing University, Nanjing,
210093, China
(Email: liuzehua@nju.edu.cn)
Qiuyin Qi a
a Department of Geography, University of Florida, Gainesville, FL, 32611, USA
(Email: qiuyinqi@ufl.edu)
* Yang Yang is the corresponding author
Tele: 1-(352)-392-3198
Fax: 1-(352)-392-8855
Yang, Y., Liu, Z-H.and Qi, Q. (2014). Domestic tourism demand of urban
and rural residents in China: Does relative income matter? Tourism
Management, 40, 193–202.
Acknowledgement: This research was financially supported by National Natural Science
Foundation of China (No. 41001070). The preliminary version of this paper was presented at 2012
TTRA International Annual Conference.
Domestic tourism demand of urban and rural residents in China:
Does relative income matter?
Abstract: The aim of this research is to investigate the domestic tourism demand of urban and
rural residents in China. Based on the data from the National Household Tourism Survey, we
specify Chinese domestic tourism demand as a function of absolute income, relative income,
domestic tourism price, and substitute price. As a major contribution of this study, relative income
is measured using the distance between individual income and average income over a
city/province. Based on the estimation results from multilevel models, this paper highlights the
effect of relative income on domestic tourism demand in some sub-regions of China. Furthermore,
regional differences between residents in different sub-regions and different patterns of
determinants between urban and rural residents are identified and discussed.
Keywords: domestic tourism demand, China, multilevel model, relative income
1. Introduction
The tourism industry boomed in China following the “reform and opening up” policy instituted in
1978. The initial incentives for tourism development were based on political and economic
considerations, and inbound tourism was given priority in China and treated as the backbone of
the tourism industry for a substantial period. Therefore, little attention had been focused on the
development of domestic tourism. However, over the past decade, rapid economic growth has
contributed to the improvement of living conditions and real growth in the income of Chinese
citizens, thereby promoting domestic travel. According to the statistics from China National
Tourism Administration, domestic tourist arrivals in China increased from 240 million in 1985 to
1,610 million in 2007. During the same period, domestic tourism receipts increased from 8 billion
RMB to 777 billion RMB, with an average annual growth rate of 23.12% (CNTA, 2008). In 1999,
the Golden Weeks—long public holidays encompassing Labour Day, National Day, and the Spring
Festival—were introduced in China to stimulate domestic travel. These long public holidays
strongly spurred the growth of China’s domestic tourism because they provided additional leisure
time for travelling to both short-distance and long-distance destinations. In 2007, 417 million
domestic tourists travelled during the three Golden Weeks, and the overall tourism receipts added
up to 182 billion RMB, accounting for 23.37% of the total domestic tourism receipts in that year
(CNTA, 2008).
Together with the rapid growth of Chinese domestic tourism, an increasing demand exists for
tourism literature in this field for policy and marketing suggestions. Using a sociological approach,
Wang (2004) proposed a theoretical model to understand the factors that contribute to tourism
consumption, including social stratification, policy change, and the marketisation of the economy.
Another paper by Wu, Zhu, and Xu (2000) identified three major factors that promote domestic
tourism in China: income growth, leisure increase, and structural adjustment of the national
economy. Using spatial analysis tools, Yang and Wong (2012a) found that a high disposable
income level and a strong propensity to travel among residents might contribute to the prosperity
of certain domestic tourism hotspots. Among a handful of studies that estimate demand for
domestic tourism in China, certain determinants have been identified empirically, including
income (Cai, Hu, & Feng, 2001; Cai & Knutson, 1998; Gu & Liu, 2004; Z. Wang, 2010),
infrastructure (Z. Wang, 2010), leisure time (Cai & Knutson, 1998), and the effect of special
economic zones (Cai, et al., 2001). However, these studies have overlooked the effects of price on
domestic tourism demand and have not considered the dichotomy of domestic tourism demand
between urban and rural residents.
In past tourism demand research, personal disposable income (which represents the absolute
income of each individual) was used as the dominant measure of the income effect (Lim, 1997).
However, tourism demand research has not taken relative income into account. Although certain
studies have advocated the inclusion of relative income in tourism demand modelling (Sauran,
1978), to the best of our knowledge, no empirical study has yet adopted this approach. Relative
income, or personal income with respect to a certain benchmark, tends to affect domestic tourism
demand because implicit income comparison affects individual economic decision-making (Cole,
Mailath, & Postlewaite, 1992, 1995). Moreover, relative income can be treated as a proxy for the
socio-economic status of each individual (Coleman, 1960). As documented by many previous
articles, socio-economic status/class influences people’s attitudes towards tourism, tourism
behaviour and expenditures on tourism activities (Moeran, 1983; Mok & Defranco, 2000; Song,
Peter, & Liu, 2000). Therefore, it is reasonable to assume that relative income should be an
important determinant of domestic tourism demand.
This paper contributes to the current body of tourism demand literature in three major ways. First,
although a few studies have attempted to consider the relative income effect on tourism, this paper
represents one of the first attempts to quantify this effect using an empirical model. By including
this variable, we expect to capture the influence of implicit income comparison on tourism
demand in the sense that tourism demand also depends ono the gap between the individual’s actual
income and selected benchmarks. Because tourism demand research has been criticised for
lacking the inclusion of non-economic factors, our research represents an important attempt in
investigating this sociological/psychological variable within tourism demand analysis. Second,
this study applies a multilevel model to analyse tourism demand under a rigorous tourism demand
analysis framework, and the model both captures the hierarchical structure of our dataset and
allows for slope heterogeneity over different areas. The results from the models discussed in this
paper could aid both the governmental and private tourism sectors in understanding the domestic
tourism demand of Chinese residents, provide insights into resource allocation to satisfy residents’
tourism demand, and consider income boom and pricing strategies to maximise profits. Third,
because the urban-rural dichotomy induces different tourism demands for urban and rural
residents (Gu & Liu, 2004), by comparing the results from models of urban and rural residents,
practitioners could be able to carry out more specific tourism planning and marketing strategies
aimed towards distinct segments of domestic tourists.
This rest of this paper is organised as follows. Section 2 discusses the research hypotheses adopted
in this study to investigate domestic tourism demand in China. Section 3 describes the data
sources and models used in this study, and Section 4 presents and explains the estimation results.
Finally, Section 5 presents several conclusions and implications based on the findings of this
study.
2. Research Hypotheses
After reviewing the previous literature on domestic tourism demand analysis, tourism marketing
in China, and sociological analysis of tourism consumption, we propose several research
hypotheses regarding the Chinese domestic tourism demand model.
An analysis and understanding of tourism demand is necessary for increasing our knowledge of
the relative importance of diverse economic determinants (Cooper, 2003). Guided by the
traditional demand theory, domestic tourism demand can be specified as a function of disposable
income, tourism price, and substitute price (Allen, Yap, & Shareef, 2009; Hamal, 1996; Seddighi
& Shearing, 1997). Income has been identified as a crucial determinant of domestic tourism
demand, which is consistent with the fact that domestic tourism is a “normal” commodity. Wang
(2010) established a VAR model to analyse Chinese domestic tourist arrivals and found income to
be an important factor. Cai and Knutson (1998) modelled Chinese domestic personal trips and
reported that GNP was a significant factor. Furthermore, Cai, Hu, and Feng (2001) used a
cross-sectional sample of thirty-five cities to study domestic tourism demand in China, and the
income elasticity was estimated to be 0.30. Another empirical paper by Gu and Liu (2004)
investigated the relationship between domestic tourism demand and household income and found
that income was the major determinant of Chinese domestic tourism demand. According to the
results from previous studies, the first hypothesis is specified as follows:
Hypothesis 1: Personal absolute income has a positive influence on domestic tourism demand in
China.
From a further review of domestic tourism demand studies, considerable variations have been
observed in the estimated income elasticities across various countries. Although certain studies
have confirmed the positive effect of income on domestic tourism demand (Garín-Muñoz, 2009;
Roget & Rodríguez González, 2006; Seddighi & Shearing, 1997; Taylor & Ortiz, 2009), other
studies have reported contradictory evidence. Salman, Shukur, and von Bergmann-Winberg (2007)
investigated the domestic tourism demand function of Swedish tourists and suggested that real
income was not of great significance. In a study on Australian domestic tourism demand,
Athanasopoulos and Hyndman (2008) found that income growth was negatively correlated; the
authors concluded that as income increases, a greater number of citizens are likely to travel abroad
instead of domestically. This negative impact of income on Australian tourism demand was also
confirmed by Allen, et al. (2009) through co-integration analysis; this group suggested that the
coefficient of income levels could be negative in the long run. Taken together, these findings
suggest the heterogeneity of the absolute income effect, which varies across different research
areas. To test this heterogeneity, we propose the following hypothesis in addition to Hypothesis 1:
Hypothesis 1a: The effect of absolute income varies across different cities/provinces in China.
As stated by the traditional demand theory, the own price of domestic tourism is expected to exert
a negative effect on domestic tourism demand, whereas the substitute price has a positive effect
(Song & Li, 2008). According to the domestic tourism literature, domestic tourism prices have
been measured in different ways. Certain studies have applied a single measurement, i.e., the
overall consumer price index (CPI) (Salman, et al., 2007), the relative CPI or other price indices
relative to an origin (Garín-Muñoz, 2009; Quayson & Var, 1982; Seddighi & Shearing, 1997), and
the price index for domestic holiday travel and accommodation (Athanasopoulos & Hyndman,
2008; Hamal, 1996), whereas others have applied more than one price variable to capture the price
effects of different components on domestic tourism (Allen, et al., 2009; Roget & Rodríguez
González, 2006). To measure the substitute price, most studies have specified the price index of
outbound tourism (Allen, et al., 2009; Hamal, 1996). However, among studies on Chinese
domestic tourism demand, no known research has incorporated price measures into empirical
models. To fill this research gap, we propose two hypotheses with respect to the effects of price
factors on Chinese domestic tourism demand:
Hypothesis 2: The domestic tourism price has a negative influence on domestic tourism demand
in China.
Hypothesis 3: The substitute price for domestic tourism has a positive influence on domestic
tourism demand in China.
Apart from the absolute income of individual residents, relative income also tends to influence
tourism demand, an observation that has been overlooked in the previous literature. Relative
income refers to personal income relative to a benchmark, i.e., the average income in a
society/country (Alpizar, Carlsson, & Johansson-Stenman, 2005; Clark, Frijters, & Shields, 2008),
and it reflects an individual’s perceived income relative to others’ income. By incorporating
relative wealth into the utility function, several authors have argued that people evaluate the
relative standing of their own income when making economic decisions (Cole, et al., 1992, 1995),
and other people’s income influences individuals’ utility via an implicit income comparison. To
the best of our knowledge, nearly all tourism demand research has overlooked the effect of
relative income and has employed only absolute income in empirical studies. Several social and
economic theories can be further extended to explain how relative income influences domestic
tourism demand, such as the theories of conspicuous consumption and individual
well-being/happiness.
Conspicuous consumption is a type of consumption designed to signal the social position and
wealth status of an individual (Veblen, 1967). Rege (2008) argued that the major incentive to
signal wealth is to gain preferential treatment via social contacts with high relative wealth. With
this in mind, tourism can communicate socio-economic status because it could be associated with
higher personal income and additional leisure time (Guo, Kim, & Timothy, 2007; Todd, 2001).
People often make their tourism consumption visible to others via pictures, souvenirs, and verbal
descriptions. To symbolise socio-economic status through tourism, tourists often purchase luxury
products (Park, Reisinger, & Noh, 2010) and services or consume fancy local foods (Y. G. Kim,
Eves, & Scarles, 2009). Many empirical studies have highlighted the influence of socio-economic
status on tourism, such as outbound travel behaviour (Moeran, 1983), the vacation decision
process (van Raaij & Francken, 1984), the tourism experience (Prentice, Witt, & Hamer, 1998),
and cultural tourism participation (H. Kim, Cheng, & O’Leary, 2007). Moreover, the influence of
socio-economic status on conspicuous tourism consumption might be more significant in China.
As indicated by Mok and Defranco (2000), in Chinese culture, the symbolic value of goods and
services is of great importance, and people value status symbols as being necessary to their daily
lives.
A large body of literature has highlighted the positive association between individual
well-being/happiness and relative income (Clark, et al., 2008; Cummins, 2000; Ferrer-i-Carbonell,
2005). People with a relatively high standing in the income hierarchy are more likely to report
greater well-being/happiness (Easterlin, 2001). It has also been suggested that people with a
higher level of well-being/happiness are more likely to participate in various tourism activities.
Gilbert and Abdullah (2004) found that people in a holiday-taking group exhibited a higher sense
of well-being before a holiday than those in a corresponding non-holiday-taking group. Therefore,
people who possess a positive attitude towards life are more likely to participate in domestic
tourism, which is a type of activity that can enhance people’s sense of happiness and achievement.
Following this logic, we argue that relative income can influence tourism demand indirectly
through its effect on individual well-being/happiness.
Therefore, we propose the following hypothesis with respect to the relationship between relative
income and domestic tourism demand:
Hypothesis 4: Relative income has a positive influence on domestic tourism demand in China.
As reported in many empirical studies, income comparisons are not symmetric (Clark & Senik,
2010; Ferrer-i-Carbonell, 2005; McBride, 2001), suggesting that the relative income effect varies
across different income groups. In the context of domestic tourism demand, the impact of relative
income is quite likely to depend on the absolute income of each individual. For those residents
with high absolute incomes, relative income tends to play a more important role in determining
domestic tourism demand to signal their wealth status. We propose the following hypothesis to
incorporate this concern:
Hypothesis 4a: The effect of relative income varies across residents in different income groups.
With China’s vast land area, great regional differences are also observed across different
sub-regions in terms of physical, cultural, and economic conditions. Therefore, regional
heterogeneity in terms of tourism demand could be significant across different sub-regions of
China. A study by Yang and Wong (2012b) highlighted significant regional differences in tourist
flow models for the eastern, central, and western parts of China. To accommodate this
heterogeneity, the authors developed separate models for each sub-region. We propose the
following hypothesis with respect to the regional differences in the Chinese domestic tourism
demand model:
Hypothesis 5: Significant differences exist among the domestic tourism demand models across
different sub-regions of China.
The hukou system refers to the residency registration system used by the Chinese government to
minimise rural-to-urban migration. Based on their hukou types, Chinese citizens are divided into
urban and rural categories, which are two apparently heterogeneous groups in terms of social
welfare and economic opportunities. Due to the existence of the hukou system and the urban-rural
dichotomy, in terms of social and economic conditions, urban and rural residents display different
consumption patterns (Yusuf, Brooks, & Zhao, 2008) as well as price and brand perceptions (Sun
& Wu, 2004). Therefore, the domestic tourism demand functions of urban and rural residents
might be different to predict (N. Wang, 2004), and we propose the following hypothesis:
Hypothesis 6: Substantial urban-rural differences exist in domestic tourism demand in China.
3. Model Specification and Data Description
The multilevel model (also known as hierarchical linear model, mixed effect model, and
contextual effects model) is specifically designed for the analysis of hierarchical data, and it
allows for the disentanglement of factors among different levels with reliable statistical results. A
simple two-level multilevel model can be specified as follows:
( ) ( ) ( )
11
++
pq
m m n n
i j t j i j t jt j i j t
mn
y x z
(1)
0
m m m
jj
(2)
where i indicates the level-1 unit (i.e., an individual resident), j indicates the level-2 unit to which
the level-1 unit belongs, and t indicates time. Equation 1 is the individual level equation, whereas
Equation 2 is the slope equation that explains the slope heterogeneity across the level-2 units. In
Equation 1, y is the dependent variable of individual i nested in the level-2 unit j at time t, α is a
constant, x is a set of p explanatory variables on level-1, z is a set of q explanatory variables on
level-2,
j
is the random effect of the level-2 unit j used to capture unobserved characteristics of
the unit, and
()i j t
is the usual error term. As indicated in Equation 1, the coefficients of the
level-1 explanatory variables
m
j
are allowed to vary across different level-2 units. In Equation
2, we define these coefficients as a mean slope
0
m
plus a random effect
m
j
that contributes to
the variation of the coefficient over different level-2 units. In practice, we allow only a subset of
the coefficients
m
j
to vary. For those fixed coefficients, we assume that the variance of error
term
m
j
is zero.
Two issues should be considered for this seemingly panel structure in Equations 1 and 2. First, a
set of time dummy variables should be included to capture the fixed time effects, which account
for the temporal changes that are the same for all level-1 and level-2 units. Second, because we
specify a random effect for each level-2 unit, this implies that the unobservable level-2
characteristics are not correlated with level-2 explanatory variables,
n
jt
z
. However, this strong
assumption is likely to be violated in practice, and a widely used solution is to reformulate
j
by
Mundlak’s (1978) formula (Baltagi, 2005), which is specified as
1
qnn
j j j
nz
(3)
In Equation 3, the level-2 random effect
j
is decomposed into two parts. The first part captures
the correlation with level-2 explanatory variables
n
jt
z
by including a linear combination of the
average over time, which is specified as
n
j
z
. In the second part of Equation 3,
j
, captures the
pure random effect, which is uncorrelated with other explanatory variables. The coefficients
n
are used specifically to correct for possible correlations, and no practical meaning commonly
exists for these coefficients. By including fixed time effects (
t
D
) and substituting Equation 3 into
Equation 1, the model becomes
( ) ( ) ( )
1 1 1
++
p q q
m m n n n n
i j t j i j t jt j t j i j t
m n n
y x z z D
(4)
We apply the multilevel model to analyse domestic tourism demand in China, covering a sample
of urban residents in thirty-five major cities from 1996 to 2007 and rural residents in thirty
provinces from 2000 to 2007. As such, we are able to identify two sources of random variation in
our domestic tourism demand model: within- and between-city/province variation. To construct
the multilevel model of domestic tourism demand, we conceptualise two levels, which are the
resident individuals (level-1) nested within the cities/provinces in which they live (level-2). Our
domestic tourism demand data are taken from the National Household Tourism Survey, obtained
from the China Domestic Tourist Survey Yearbook (1997-2008). This dataset is aggregated by
income groups in different sample cities/provinces. To disaggregate this data, we weight each
income group by the corresponding number of observations. Therefore, the only level-1
information we can obtain is the income group to which each resident belongs, and the income
variable is available over intervals instead of on a continuous scale. In previous Chinese tourism
demand studies, tourism demand data were aggregated by administrative units, and this imposes
the restriction of full homogeneity of each individual within the unit in terms of economic status.
If we ignore individual heterogeneity, we run the risk of incurring an ecological or aggregation
fallacy (Robinson, 1950), which suggests that the results from group-level studies might not be
valid when transferred to the individual level. Therefore, to draw a more reliable conclusion, it is
important to take personal income information into consideration in domestic tourism demand
analysis.
A set of explanatory variables in different levels is explained in Table 1. The dependent variable in
the model is lnD, which is the annual domestic tourism expenditure per person (in log). For the
level-1 explanatory variables, because personal income data are banded into seven categories, a
set of dummy variables, inc(2) to inc(7), are used to capture income effects, leaving inc(1) as a
reference. Moreover, the variable RI is introduced to capture the relative income effect, which is
defined as the log of personal income minus the log of the average income of the city/province.
Therefore, we assume that domestic tourism demand depends on the distance between an
individual’s own income and the average income of the city/province. We calculate the personal
income in terms of the mid-point of each income group and assign a value of half of the upper
bound for the lowest income group and 1.5 times the lower bound for the highest income group.
The data for the average income of each city/province were obtained from the China Statistical
Yearbook (1997-2008).
With respect to other explanatory variables, lnP is the tourism price index (in log) and lnPS is the
substitute price index for tourism (in log). We treat lnP and lnPS as level-2 variables by assuming
that residents in the same city/province in the same year face the same price and substitution price
for domestic tourism. The tourism price index is constructed following the origin-destination
matrix weighted method developed by Lanza, Temple, and Urga (2003). The tourism price index
P of level-2 unit j at time t is specified as
ˆ
jkt
jt kt
kjkt
k
TA
PP
TA
(5)
where TAjkt is the number of tourist arrivals from origin j to destination k at time t, the sum
jkt
kTA
is the total number of tourist arrivals from origin j at time t, and
ˆkt
P
is a measure of
the price index in destination k at time t. We use CPI (2000 = 100) as the price index measurement
for urban residents and use the price index of tourism (2000 = 100) for rural residents because we
cannot obtain the price index of tourism from 1996-1999 for urban residents, and the preliminary
results suggest that CPI and the price index of tourism are highly correlated. In terms of the
substitute price, we cannot specify outbound tourism as a substitute good for domestic tourism in
the demand function (Allen, et al., 2009; Hamal, 1996) due to visa restrictions and the relatively
low GDP per capita for Chinese residents in general. Instead, we use the local price index of
cultural activities and entertainment (2000 = 100) by assuming that local cultural activities and
entertainment are a substitute good for domestic tourism. City/province level price data were
obtained from the China Price Statistical Yearbook (1997-2008), and tourist arrival data were
obtained from the China Domestic Tourist Survey Yearbook (1997-2008). As indicated in
Equation 4, we also include a set of year dummy variables to capture the fixed time effects, such
as the impact of the SARS outbreak in 2003. Moreover, following Mundlak’s formula (Equations
3 and 4), we include three time-invariant explanatory variables to alleviate the possible
endogeneity problem of the level-2 random effect, which include the average city/province income
(in log), the average tourism price index (in log), and the average substitute price index for
tourism (in log) of the city/province over the study period.
There are several noticeable advantages of applying multilevel models in this study. First, the
multilevel model takes full advantage of the hierarchical structure of our dataset to avoid the
ecological fallacy induced by ignoring level-1 information. Second and more importantly, the
multilevel model accounts for the heterogeneity across different levels and specifies the random
effects that occur over particular levels. Equation 4 incorporates the random effect
j
of each
city/province j, which is used to capture the unobserved city/province specific factors that
influence tourism demand but have not been incorporated in our model, such as location relative
to major tourist attractions and travelling cultures. In the slope equation (Equation 2), a set of
m
j
values can also be used to capture the variation of slopes across different cities/provinces (level-2
units).
A large number of models could be estimated to consider all possible random effects and
interaction terms. However, because we are particularly interested in various income effects on
Chinese domestic tourism demand, only the models relevant to our research goals are estimated. A
three-step sequential modelling strategy is adopted to successively introduce complexity. The first
model uses Equations 2 and 4 and only captures the absolute income effect, excluding the
explanatory variable RI; this model is labelled the “baseline model”. In the second model, in
addition to the baseline model, RI is introduced to capture the relative income effect over
cities/provinces. In the specified model, a positive estimated coefficient of RI indicates the
expected relative income effect. For example, imagine that there are two people with the same
personal income in two different level-2 units (different cities/provinces). The one living in the
unit with a lower average income exhibits a higher relative income, which contributes to a higher
level of domestic tourism demand given the same tourism price and substitute price. In the third
model, we test the asymmetry of the relative income effect in different income groups (Hypothesis
4a). Therefore, the explanatory variable RI is replaced by seven interaction terms, i.e., inc(1)*RI to
inc(7)*RI, to estimate the specific relative income effect of each income group.
Several estimation issues should be noted. We use the full maximum likelihood estimation (FMLE)
with expectation-maximisation (EM) iterations to estimate the specified multilevel model. The
FMLE generates robust estimates with a large sample size, and the estimates are asymptotically
efficient and consistent (Hox, 2010). Because Equations 2 and 4 include several random
components, we assume that these components are independent from each other and follow
normal distributions with a zero mean and finite variance. Therefore, their co-variances are set to
zero to avoid the burdensome co-variance structure in computation. Furthermore, although the
random effect cannot be directly estimated during the maximum likelihood estimation, we can
obtain the estimate from best linear unbiased predictions (BLUPs) (Bates & Pinheiro, 1998).
Finally, we compute the robust standard error of each coefficient from the clustered variance
calculated from level-2 (city/province level) in the multilevel models.
The descriptive statistics for the variables specified are presented in Table 1. The sample consists
of 138,797 urban residents and 40,840 rural residents in China. A proportion of 54.04% of urban
residents fall into income groups 3 and 4, with a monthly income between 1,000 and 2,999 RMB
Yuan during the study period. Among the rural residents, the sample is relatively evenly
distributed across different income groups. The mean values of RI are 0.939 and -0.064 for urban
and rural residents, respectively, suggesting that in the sample, a larger number of urban residents
have a personal income that is higher than the average income of the city/province. The average
values of lnP and lnPS are 4.626 and 4.632 for urban residents, respectively, and 4.589 and 4.627
for rural residents, respectively.
(Insert Table 1 here)
4. Results
4.1. Demand model for urban residents
Table 2 presents the estimation results of domestic tourism demand models for urban residents.
The first three models in the table, Urban-All-1, Urban-All-2, and Urban-All-3, include all
138,797 observations across thirty five cities. In the Urban-All-1 model, which is the baseline
model, only the absolute income effect is incorporated. The estimated coefficients of inc(2) to
inc(7) are positive, statistically significant, and in ascending order, suggesting a positive and
significant absolute income effect on the domestic tourism demand of Chinese urban residents.
lnPS is estimated to be positive and significant, whereas lnP is insignificant. In the Urban-All-2
model incorporating the relative income effect, RI is estimated to be insignificant albeit positive.
Moreover, in the Urban-All-3 model, after including a set of interaction terms to capture the
asymmetry of the relative income effect, these interaction terms of RI are still not significant.
Therefore, based on nationwide data, the influence of relative income on domestic tourism
demand of urban residents is limited and insignificant.
(Insert Table 2 here)
Table 2 also presents the estimation results of various random effects, which are the estimated
variances of random effects. In the multilevel model, a common way to test the significance of
random effects is the Likelihood Ratio (LR) test (Rabe-Hesketh & Skrondal, 2008). As suggested
by the results of LR tests in Table 2, all random effects are estimated to be statistically significant
in the first three models, highlighting the substantial heterogeneity across cities. From the
estimated variance of each random effect, one can compare the degree of the heterogeneity. For
example, the estimated variances of the random effects for high-income residents, like VAR(inc(6))
and VAR(inc(7)), are larger than their counterparts of low-income residents, suggesting a more
intense cross-city heterogeneity in absolute income effects for high-income residents. Furthermore,
we can predict this random effect and unveil the absolute income effect for each city. Figure 1
illustrates the average predicted random effect of absolute income through BLUPs. Each circle
represents each sampled city, and its size reflects the magnitude of this average effect. As shown
in the map, one noticeable finding is that in some of the most developed cities, like those in the
Yangtze River Delta and the Pearl River Delta, this average effect is moderate, suggesting a
modest absolute income effect on urban residents’ domestic tourism demand in the developed
area.
(Insert Fig. 1 here)
To compare the absolute income effect for each geographic sub-region, we use the same
specification as the Urban-All-1 model to estimate the dataset of each sub-region. A common
approach to regionalising China is to divide it into eastern, central, and western regions (Yang &
Wong, 2012b). Due to space limitations, we do not present the detailed estimation results in this
work. Instead, to compare the estimates intuitively, Figure 2 shows the estimated coefficients of
inc(2) to inc(7) for the different sub-regions and illustrates that the absolute income effect is
always greatest for the western urban residents and smallest for the eastern urban residents. This
result suggests that in the least developed areas, i.e., the west of China, absolute income plays a
more important role in determining the level of domestic tourism demand.
(Insert Fig. 2 here)
We estimate the models in the last four columns of Table 2, including the relative income effect
with the sample in different sub-regions. From the Urban-East-2, Urban-Centre-2, and
Urban-West-2 models, we find that between the two price variables, only lnPS is statistically
significant for the east and centre and is estimated to be positive. RI is estimated to be significant
and positive for urban residents in eastern cities. The positive coefficient suggests that, depending
on personal absolute income, a higher relative income is associated with a higher level of domestic
tourism demand for urban residents in the east of China. This result supports Hypothesis 4 and
highlights the importance of implicit income comparison in determining domestic tourism demand.
Not only absolute income but also relative income is likely to influence individual tourism
demand. To further investigate the magnitude of this relative income effect for different income
groups, we developed the Urban-East-3 model, which includes a set of interaction terms between
the income dummies and relative income. This model shows that inc(4)*RI, inc(5)*RI, and
inc(6)*RI are estimated to be positive and significant. This result highlights the significant relative
income effect for middle- and high-income residents (monthly income between 2,000 and 4,999
RMB Yuan) and is also consistent with the findings reported by McBride (2001) in that the
relative income effects are smaller at low-income levels. To test the heterogeneity of the relative
income effect over different income groups (Hypothesis 4a), we use the Wald test to validate the
equality of seven interaction terms. The test statistic is 29.47 with six degrees of freedom, thus
rejecting the null hypothesis of homogeneity at the 0.01 significance level.
4.2. Demand model for rural residents
In this section, we use the same methods to examine the domestic tourism demand of rural
residents. Table 3 presents the estimation results of models for Chinese rural residents across thirty
provinces from 2000 to 2007. The Rural-All-1 model is the baseline model that incorporates
nationwide data. Variables inc(2) to inc(7) are estimated to be positive, significant, and in
ascending order, emphasising the positive effect of absolute income on the domestic tourism
demand of Chinese rural residents. Neither lnP nor lnPS are statistically significant, although their
signs are consistent with the traditional demand theory. The estimated substitute elasticity is 0.533
in the Rural-All-1 model and is lower than the estimate in the Urban-All-1 model, which is 1.252.
This result suggests that the substitution effect of local cultural activities and entertainment for
domestic tourism is stronger for urban residents. The Rural-All-2 and Rural-All-3 models
incorporate symmetric and asymmetric relative income effects, respectively. RI and all interaction
terms of the income dummies are estimated to be insignificant in these two models. Similarly to
the results of various urban models, the nationwide data do not support the significant relative
income effect on the domestic tourism demand of Chinese rural residents.
(Insert Table 3 here)
With respect to the slope heterogeneity in the Urban-All-2 model, LR tests suggest that all random
effects are statistically significant. This result indicates that the absolute income effects on rural
residents’ domestic tourism demand are also heterogeneous across different provinces. The larger
estimated variances for the high-income groups highlight the more intense cross-province
heterogeneity in absolute income effects. We can predict the random effects of each province
using BLUPs. Figure 3 shows the average predicted random effects of the income dummy
variables, which are calculated in the same way as those shown in Figure 1. This average effect is
found to be strongest in eastern provinces, such as Fujian and Shandong.
(Insert Fig. 3 here)
To compare the absolute income effect from the eastern, central, and western sub-samples, Figure
4 presents the results of baseline models for different sub-regions, which are similar to those
presented in Figure 2. The absolute income effects in the east are always greater than those in the
centre. However, for the west, we observe that the absolute income effects are greater than those
of other regions for lower-income groups and smaller for higher-income groups.
(Insert Fig. 4 here)
In the last four models presented in Table 3, we compare the estimation results, including the
relative income effects for different sub-regions. In the Rural-West-2 model, lnP is estimated to be
significant, suggesting that the own price effect is significant for western rural residents in China.
Moreover, the significant and positive estimated coefficient of RI in the Rural-Centre-2 model
highlights the significant relative income effect for rural residents in central provinces. By
incorporating a set of interaction terms to capture the possible asymmetry of this relative income
effect, the Rural-Centre-3 model indicates that the relative income effect is greater for
middle-income rural residents in central provinces. A Wald test statistic of 158.23 with six degrees
of freedom rejects the null hypothesis of the homogeneous relative income effect across different
income groups in central provinces.
4.3. Summary of results
After obtaining the estimation results of the multilevel models, we summarise the conclusions
obtained for the research hypotheses proposed in Section 2. We cannot reject Hypothesis 1 and
identify significant and dominant absolute income effects on domestic tourism demand for both
urban and rural residents. This result is consistent with the findings reported by Cai, et al. (2001),
Gu and Liu (2004), and Wang (2010). Therefore, it is projected that, together with China’s
economic growth and the concomitant increase in household income expected in the future,
Chinese domestic tourism demand will continue to increase. Moreover, the cross-city/province
heterogeneity of this absolute income effect was identified, and Hypothesis 1a was accepted. We
found that this effect is stronger in certain cities/provinces than in others. More importantly, this
heterogeneity was discovered to be more intense for high-income groups. With respect to price
effects, because lnP was estimated to be insignificant in most models, the results did not support
Hypothesis 2, which predicts a negative own price effect on domestic tourism demand. However,
the significant and positive coefficient of lnPS in the urban models supports Hypotheses 3,
predicting a positive substitute price effect for Chinese urban residents.
With respect to the relative income effect, Hypothesis 4 is not supported by the nationwide data;
however, we did find evidence to support this hypothesis with sub-regional data. For instance, the
relative income effect was found to be significant for eastern urban residents and central rural
residents, suggesting that the richer an individual is compared with others within the same
city/province, the higher the level of domestic tourism demand that individual will display.
Moreover, depending on which sub-regions exhibited a significant relative income effect, we
found that this effect varies across different income groups, which supports Hypothesis 4a. The
results showed that the relative income effect is less intense for lower-income residents. Finally,
we identified substantial differences across residents in different sub-regions and between urban
and rural residents, and Hypotheses 5 and 6 are thus corroborated. For instance, the results
suggested that absolute income plays a more important role in determining domestic tourism
demand for eastern urban residents than in determining that in other sub-regions, and the
substitute price effect is more substantial for urban residents than rural ones.
5. Conclusion
This study applied multilevel models to investigate the domestic tourism demand of urban and
rural residents in China. The data from the National Household Tourism Survey covered urban
residents in thirty-five major cities from 1996 to 2007 and rural residents in thirty provinces from
2000 to 2007. Absolute personal income was found to be the dominant factor that influenced
Chinese domestic tourism demand for both urban and rural residents. According to the results of
the multilevel models, this absolute income effect varies across different cities/provinces, showing
significant heterogeneity. Moreover, this paper breaks new ground by estimating the effect of
relative income on domestic tourism demand and highlights the significant relative income effect
on tourism demand in certain sub-regions of China. For those sub-regions with significant relative
income effects, we found that this effect is asymmetric and is smaller for low-income groups.
Based on these findings, several insights are provided in terms of government policy and
marketing strategy. First, when designing marketing plans to target potential tourists, relative
income should be another important factor to consider apart from absolute income because in
certain areas, it also determines the level of domestic tourism demand. Second, different
marketing strategies should be proposed for residents in different areas as well as residents in
urban and rural areas. For example, urban residents in the east are more concerned with relative
income, and western urban residents are more sensitive to absolute income. Third, depending on
the fixed price, tourism products and services should be designed with additional status signalling
to satisfy the needs of certain residents because domestic tourism is a type of conspicuous
consumption for such residents.
Certain limitations of this research should be noted. We could not obtain the micro-data for
individual residents, and the data actually consist of the weighted data of each income group
nested in each sample city/province. Therefore, the socio-demographic information of each
individual cannot be obtained. Moreover, our model is static and does not incorporate dynamic
factors. No additional information on short-term and long-term effects can be obtained.
Furthermore, because previous research has indicated that seasonality is significant in domestic
tourism demand (Deng & Athanasopoulos, 2011), future research should apply quarterly data to
model the demand in place of the annual data used in this work. Finally, although our paper
presents evidence of the relative income effect on domestic tourism demand, it is of interest to
define more accurate reference groups of income comparison rather than using the average income
of cities/provinces.
References
Allen, D., Yap, G., & Shareef, R. (2009). Modelling interstate tourism demand in Australia: A
cointegration approach. Mathematics and Computers in Simulation, 79, 2733-2740.
Alpizar, F., Carlsson, F., & Johansson-Stenman, O. (2005). How much do we care about absolute
versus relative income and consumption? Journal of Economic Behavior & Organization, 56,
405-421.
Athanasopoulos, G., & Hyndman, R. J. (2008). Modelling and forecasting Australian domestic tourism.
Tourism Management, 29, 19-31.
Baltagi, B. H. (2005). Econometric Analysis of Panel Data (3rd ed.). Chichester ; Hoboken, NJ: J.
Wiley & Sons.
Bates, D. M., & Pinheiro, J. C. (1998). Computational methods for multilevel modelling. In Technical
Memorandum BL0112140-980226-01TM. Murray Hill, NJ: Bell Labs, Lucent Technologies.
Cai, L. A., Hu, B., & Feng, R. (2001). Domestic tourism demand in China's urban centres: Empirical
analyses and marketing implications. Journal of Vacation Marketing, 8, 64-74.
Cai, L. A., & Knutson, B. J. (1998). Analyzing domestic tourism demand in China - A behavioral
model. Journal of Hospitality & Leisure Marketing, 5, 95-113.
Clark, A. E., Frijters, P., & Shields, M. A. (2008). Relative income, happiness, and utility: An
explanation for the Easterlin Paradox and other puzzles. Journal of Economic Literature, 46,
95-144.
Clark, A. E., & Senik, C. (2010). Who compares to whom? The anatomy of income comparisons in
Europe. The Economic Journal, 120, 573-594.
CNTA. (2008). The Yearbook of China Tourism Statistics 2007. Beijing: China Tourism Press.
Cole, H. L., Mailath, G. J., & Postlewaite, A. (1992). Social norms, savings behavior, and growth.
Journal of Political Economy, 100, 1092-1125.
Cole, H. L., Mailath, G. J., & Postlewaite, A. (1995). Incorporating concern for relative wealth into
economic models. Federal Reserve Bank of Minneapolis Quarterly Review, 19, 12-21.
Coleman, R. P. (1960). The significance of social stratification in selling. In M. L. Bell (Ed.),
Marketing: A Maturing Discipline (pp. 177-184). Chicago: American Marketing Association.
Cooper, C. (2003). Classic Reviews in Tourism. Clevedon: Channel View.
Cummins, R. A. (2000). Personal income and subjective well-being: A review. Journal of Happiness
Studies, 1, 133-158.
Deng, M., & Athanasopoulos, G. (2011). Modelling Australian domestic and international inbound
travel: A spatial–temporal approach. Tourism Management, 32, 1075-1084.
Easterlin, R. A. (2001). Income and happiness: Towards a unified theory. The Economic Journal, 111,
465-484.
Ferrer-i-Carbonell, A. (2005). Income and well-being: An empirical analysis of the comparison income
effect. Journal of Public Economics, 89, 997-1019.
Garín-Muñoz, T. (2009). Tourism in Galicia: Domestic and foreign demand. Tourism Economics, 15,
753-769.
Gilbert, D., & Abdullah, J. (2004). Holidaytaking and the sense of well-being. Annals of Tourism
Research, 31, 103-121.
Gu, H., & Liu, D. (2004). The relationship between resident income and domestic tourism in China.
Tourism Recreation Research, 29, 25-33.
Guo, Y., Kim, S. S., & Timothy, D. J. (2007). Development characteristics and implications of
mainland Chinese outbound tourism. Asia Pacific Journal of Tourism Research, 12, 313-332.
Hamal, K. (1996). Modeling domestic holiday tourism demand in Australia: Problems and solutions.
Asia Pacific Journal of Tourism Research, 1, 35-46.
Hox, J. J. (2010). Multilevel Analysis: Techniques and Applications (2nd ed.). New York: Routledge.
Kim, H., Cheng, C.-K., & O’Leary, J. T. (2007). Understanding participation patterns and trends in
tourism cultural attractions. Tourism Management, 28, 1366-1371.
Kim, Y. G., Eves, A., & Scarles, C. (2009). Building a model of local food consumption on trips and
holidays: A grounded theory approach. International Journal of Hospitality Management, 28,
423-431.
Lanza, A., Temple, P., & Urga, G. (2003). The implications of tourism specialisation in the long run:
An econometric analysis for 13 OECD economies. Tourism Management, 24, 315-321.
Lim, C. (1997). Review of international tourism demand models. Annals of Tourism Research, 24,
835-849.
McBride, M. (2001). Relative-income effects on subjective well-being in the cross-section. Journal of
Economic Behavior & Organization, 45, 251-278.
Moeran, B. (1983). The language of Japanese tourism. Annals of Tourism Research, 10, 93-108.
Mok, C., & Defranco, A. L. (2000). Chinese cultural values: Their implications for travel and tourism
marketing. Journal of Travel & Tourism Marketing, 8, 99-114.
Mundlak, Y. (1978). On the pooling of time series and cross section data. Econometrica, 46, 69-85.
Park, K.-S., Reisinger, Y., & Noh, E.-H. (2010). Luxury shopping in tourism. International Journal of
Tourism Research, 12, 164-178.
Prentice, R. C., Witt, S. F., & Hamer, C. (1998). Tourism as experience: The case of heritage parks.
Annals of Tourism Research, 25, 1-24.
Quayson, J., & Var, T. (1982). A tourism demand function for the Okanagan, BC. Tourism Management,
3, 108-115.
Rabe-Hesketh, S., & Skrondal, A. (2008). Multilevel and Longitudinal Modeling Using Stata (2nd ed.).
College Station, TX: Stata Press.
Rege, M. (2008). Why do people care about social status? Journal of Economic Behavior &
Organization, 66, 233-242.
Robinson, W. S. (1950). Ecological correlations and the behavior of individuals. American Sociological
Review, 15, 351-357
Roget, F. M., & Rodríguez González, X. A. (2006). Rural tourism demand in Galicia, Spain Tourism
Economics, 12, 21-31.
Salman, A. K., Shukur, G., & von Bergmann-Winberg, M.-L. (2007). Comparison of econometric
modelling of demand for domestic and international tourism: Swedish data. Current Issues in
Tourism, 10, 323-342.
Sauran, A. (1978). Economic determinants of tourist demand: A survey. Tourism Review, 3, 2-4.
Seddighi, H. R., & Shearing, D. F. (1997). The demand for tourism in North East England with special
reference to Northumbria: an empirical analysis. Tourism Management, 18, 499-511.
Song, H., & Li, G. (2008). Tourism demand modelling and forecasting—A review of recent research.
Tourism Management, 29, 203-220.
Song, H., Peter, R., & Liu, X. (2000). An empirical study of outbound tourism demand in the UK.
Applied Economics, 32, 611-624.
Sun, T., & Wu, G. (2004). Consumption patterns of Chinese urban and rural consumers. The Journal of
consumer marketing, 21, 245-253.
Taylor, T., & Ortiz, R. A. (2009). Impacts of climate change on domestic tourism in the UK: A panel
data estimation. Tourism Economics, 15, 803-812.
Todd, S. (2001). Self-concept: A tourism application. Journal of Consumer Behaviour, 1, 184-196.
van Raaij, W. F., & Francken, D. A. (1984). Vacation decisions, activities, and satisfactions. Annals of
Tourism Research, 11, 101-112.
Veblen, T. (1967). The Theory of the Leisure Class. New York: Penguin Books.
Wang, N. (2004). The rise of touristic consumerism in urban China. Tourism Recreation Research, 29,
47-58.
Wang, Z. (2010). Factors that influence the gowth of Chinese domestic tourism arrivals (1985–2007) –
An empirical research based on the VAR model. Asia Pacific Journal of Tourism Research, 15,
449-459.
Wu, B., Zhu, H., & Xu, X. (2000). Trends in China's domestic tourism development at the turn of the
century. International Journal of Contemporary Hospitality Management, 12, 296-299.
Yang, Y., & Wong, K. K. F. (2012a). Spatial distribution of tourist flows to China's cities. Tourism
Geographies, 10.1080/14616688.14612012.14675511.
Yang, Y., & Wong, K. K. F. (2012b). A spatial econometric approach to model spillover effects in
tourism flows. Journal of Travel Research, 51, 768-778.
Yusuf, F., Brooks, G., & Zhao, P. (2008). Household Consumption in China: An Examination of the
Utility of Urban-Rural Segmentation. In S. H. Murdock & D. A. Swanson (Eds.), Applied
Demography in the 21st Century (pp. 285-298). Netherlands: Springer
Table 1
Descriptive statistics of variables
Urban Resident Model
Rural Resident Model
Continuous Variable
Mean
Std.
Dev.
Mean
Std.
Dev.
lnD
Annual domestic tourism expenditure per person
(in log)
6.463
0.591
6.306
0.559
lnP
Tourism price index (in log) in Equation 5
4.626
0.039
4.589
0.078
lnPS
Substitute price index of domestic tourism (in
log)
4.632
0.088
4.627
0.057
RI
The distance between personal income (in log)
and the average city/province income (in log)
0.939
0.627
-0.064
0.525
Categorical Variable
Monthly Income
(in RMB Yuan)
Percent
Annual Income
(in RMB Yuan)
Percent
inc(1)
~499
3.12
~1,499
10.44
inc(2)
500~999
11.42
1,500~1,999
14.38
inc(3)
1,000~1,999
31.42
2,000~2,499
16.53
inc(4)
2,000~2,999
22.62
2,500~2,999
13.57
inc(5)
3,000~3,999
13.83
3,000~3,999
19.74
inc(6)
4,000~4,999
7.73
4,000~4,999
10.87
inc(7)
5,000~
9.85
5,000~
14.48
Number of Observations
Level-1 unit
138,797
40,840
Level-2 unit
35
30
Time period
1996-2007
2000-2007
Table 2
Estimation results for domestic tourism demand of urban residents
Variable
Urban-All-1
Urban-All-2
Urban-All-3
Urban-East-2
Urban-East-3
Urban-Centre
-2
Urban-West-2
inc(2)
0.407***
0.402
0.503
-0.0420
0.690**
0.792
1.062*
(0.066)
(0.337)
(0.361)
(0.252)
(0.351)
(0.533)
(0.548)
inc(3)
0.696***
0.687
0.832
0.0316
0.841*
1.284
1.653*
(0.062)
(0.518)
(0.541)
(0.322)
(0.438)
(0.872)
(0.870)
inc(4)
0.908***
0.898
0.903
0.102
0.804*
1.580
2.116*
(0.069)
(0.661)
(0.670)
(0.400)
(0.481)
(1.120)
(1.082)
inc(5)
1.040***
1.027
0.911
0.144
0.610
1.762
2.408*
(0.070)
(0.755)
(0.782)
(0.443)
(0.543)
(1.271)
(1.239)
inc(6)
1.154***
1.140
1.358
0.144
0.696
1.929
2.723**
(0.083)
(0.832)
(0.834)
(0.493)
(0.585)
(1.398)
(1.365)
inc(7)
1.276***
1.260
1.805
0.174
1.117*
2.181
2.989*
(0.095)
(0.968)
(1.115)
(0.572)
(0.614)
(1.563)
(1.636)
lnP
0.428
0.435
0.622
-1.228
-1.298
2.718
2.922
(1.586)
(1.585)
(1.626)
(2.727)
(2.805)
(1.980)
(1.950)
lnPS
1.252***
1.253***
1.294***
1.607***
1.683***
0.548**
0.558
(0.395)
(0.397)
(0.391)
(0.434)
(0.425)
(0.237)
(0.954)
RI
0.00472
0.309**
-0.281
-0.478
(0.275)
(0.149)
(0.478)
(0.455)
inc(1)*RI
-0.107
-0.308
(0.316)
(0.269)
inc(2)*RI
0.0574
0.287
(0.307)
(0.288)
inc(3)*RI
-0.0436
0.183
(0.296)
(0.192)
inc(4)*RI
0.0942
0.343**
(0.283)
(0.154)
inc(5)*RI
0.170
0.528***
(0.302)
(0.167)
inc(6)*RI
-0.0698
0.439**
(0.303)
(0.181)
inc(7)*RI
-0.215
0.193
(0.353)
(0.160)
constant
-46.72
-46.76
-45.38
-259.8***
-259.3***
-18.43
48.13
(36.080)
(36.467)
(36.236)
(60.041)
(64.594)
(55.931)
(43.956)
Random effects
VAR(city)
0.151
0.151
0.149
0.067
0.074
0.058
0.141
VAR(inc(2))
0.129
0.129
0.115
0.207
0.182
0.045
0.054
VAR(inc(3))
0.116
0.116
0.111
0.158
0.165
0.032
0.085
VAR(inc(4))
0.142
0.142
0.128
0.191
0.182
0.071
0.086
VAR(inc(5))
0.140
0.141
0.126
0.167
0.168
0.112
0.094
VAR(inc(6))
0.201
0.201
0.191
0.207
0.191
0.107
0.216
VAR(inc(7))
0.254
0.254
0.245
0.250
0.248
0.221
0.264
LR test
93588.785***
93573.895***
91473.917***
35718.087***
35924.482***
6760.732***
21727.755***
Obs.
138797(35)
138797(35)
138797(35)
57086(35)
78295(16)
19495(8)
41007(11)
AIC
128040.4
112613.7
111098.3
30944.0
41839.9
14777.1
44122.4
BIC
128168.3
112800.6
111344.3
31158.9
41978.9
14832.2
44208.7
(Notes: * indicates p<0.10, ** indicates p<0.05, *** indicates p<0.01.Robust standard error in parenthesis. The
estimates of year dummies and λ’s (in Equation 4) are not presented. VAR(.) indicates the estimated variance of
random effects, and LR test refers to the Likelihood Ratio test for random effects based on non-robust standard
error model)
Table 3
Estimation results for domestic tourism demand of rural residents
Variable
Rural-All-1
Rural -All-2
Rural -All-3
Rural -East-2
Rural
-Centre-2
Rural
-Centre-3
Rural -West-2
inc(2)
0.260***
0.398
0.614
1.244
-1.714**
-0.884
-0.118
(0.051)
(0.586)
(0.594)
(0.946)
(0.702)
(0.866)
(0.912)
inc(3)
0.309***
0.488
0.701
1.525
-2.331**
-1.567
-0.0955
(0.061)
(0.756)
(0.760)
(1.249)
(0.916)
(1.002)
(1.189)
inc(4)
0.338***
0.549
0.783
1.757
-2.785**
-1.986*
-0.129
(0.070)
(0.904)
(0.895)
(1.466)
(1.084)
(1.139)
(1.415)
inc(5)
0.443***
0.692
0.910
2.285
-3.189**
-2.448*
-0.300
(0.068)
(1.068)
(1.052)
(1.760)
(1.307)
(1.345)
(1.630)
inc(6)
0.564***
0.855
1.158
2.678
-3.654**
-2.596*
-0.267
(0.078)
(1.247)
(1.204)
(2.052)
(1.476)
(1.517)
(1.966)
inc(7)
0.798***
1.171
1.320
3.631
-4.835**
-3.927**
-0.233
(0.094)
(1.604)
(1.553)
(2.644)
(1.962)
(1.866)
(2.477)
lnP
-0.307
-0.328
-0.335
0.526
0.762
0.636
-0.473**
(0.391)
(0.389)
(0.381)
(1.099)
(1.308)
(1.285)
(0.188)
lnPS
0.553
0.518
0.559
0.218
-2.685**
-2.509*
1.662
(0.628)
(0.597)
(0.636)
(0.809)
(1.316)
(1.338)
(1.504)
RI
-0.163
-1.130
2.324***
0.444
(0.690)
(1.137)
(0.838)
(1.064)
inc(1)*RI
-0.351
1.665*
(0.682)
(0.878)
inc(2)*RI
-0.200
2.442***
(0.701)
(0.799)
inc(3)*RI
-0.244
2.172**
(0.666)
(0.888)
inc(4)*RI
-0.176
2.706***
(0.710)
(0.938)
inc(5)*RI
0.00741
2.729***
(0.679)
(0.804)
inc(6)*RI
-0.378
1.512*
(0.624)
(0.867)
inc(7)*RI
-0.0543
2.178***
(0.677)
(0.786)
constant
-6.779
-5.662
-5.664
3.903
-1.883
0.461
-19.45
(9.122)
(9.946)
(9.883)
(11.758)
(20.181)
(18.403)
(34.335)
Random effects
VAR(provin
ce)
0.097
0.097
0.097
0.090
0.030
0.023
0.123
VAR(inc(2))
0.058
0.058
0.058
0.130
0.060
0.062
0.017
VAR(inc(3))
0.077
0.077
0.075
0.172
0.032
0.030
0.035
VAR(inc(4))
0.108
0.108
0.103
0.205
0.011
0.013
0.084
VAR(inc(5))
0.095
0.095
0.116
0.122
0.008
0.015
0.144
VAR(inc(6))
0.126
0.127
0.128
0.157
0.036
0.028
0.183
VAR(inc(7))
0.190
0.191
0.210
0.241
0.097
0.104
0.165
LR test
15412.953***
15416.130***
15394.173***
3804.368***
2693.188***
2817.484***
5792.902 ***
Obs.
40840
40840
40840
13259
13641
13641
13940
AIC
33937.6
33934.9
33759.1
10710.5
6647.0
5893.8
14451.9
BIC
34170.3
34176.2
34009.0
10785.4
6699.6
5946.5
14527.3
(Notes: * indicates p<0.10, ** indicates p<0.05, *** indicates p<0.01.Robust standard error in parenthesis. The
estimates of year dummies and λ’s (in Equation 4) are not presented. VAR(.) indicates the estimated variance of
random effects, and LR test refers to the Likelihood Ratio test for random effects based on non-robust standard
error model)
Fig. 1. Estimated random effects of urban residents’ income for cities
Fig. 2. Estimated coefficient of absolute income on urban residents’ domestic tourism demand for different
sub-regions
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
inc(1) inc(2) inc(3) inc(4) inc(5) inc(6) inc(7)
Estimated Coefficient
Income Group
East
Center
West
Fig. 3. Estimated random effects of rural residents’ income for provinces
Fig. 4. Estimated coefficient of absolute income on rural residents’ domestic tourism demand for different
sub-regions
0
0.2
0.4
0.6
0.8
1
1.2
inc(1) inc(2) inc(3) inc(4) inc(5) inc(6) inc(7)
Estimated Coefficient
Income Group
East
Center
West
