Explaining regional economic multipliers of tourism: does cross-regional heterogeneity exist?

Abstract

This study investigates the determinants of tourism-related economic multipliers, which are calibrated from regional input–output tables of 30 Chinese provinces. Several latent class regression models are applied to capture cross-regional heterogeneity and investigate the determinants of output, income, and employment multipliers. Results indicate that the level of economic development is positively associated with output multiplier, and it is also positively associated with employment multiplier when the multiplier is large. Moreover, the size of tourism economy is negatively associated with employment multiplier when the multiplier is small. Some policy implications are presented based on empirical results.

Full text

EXPLAINING REGIONAL ECONOMIC MULTIPLIERS OF TOURISM:
DOES CROSS-REGIONAL HETEROGENEITY EXIST?
Yang Yang* Timothy J. Fik** Benjamin Altschuler***
* School of Sport, Tourism and Hospitality Management, Temple University,
Philadelphia, PA, 19122, United States
Phone: (01) 215-204-8701; Fax: (01) 215-204-8705
Email: yangy@temple.edu
** Department of Geography, University of Florida
E-mail: fik@ufl.edu
***School of Sport, Tourism and Hospitality Management, Temple University,
Email: benjamin.altschuler@temple.edu
Corresponding Author: Dr. Yang Yang
Please cite as:
Yang, Y., Timothy, F. J., and Altschuler, B. (2018). Explaining regional economic
multipliers of tourism: does cross-regional heterogeneity exist? Asia Pacific Journal
of Tourism Research, 23(1): 15-23.

1
EXPLAINING REGIONAL ECONOMIC MULTIPLIERS OF TOURISM: DOES
CROSS-REGIONAL HETEROGENEITY EXIST?
Abstract: This study investigates the determinants of tourism-related economic multipliers,
which are calibrated from regional input-output tables of 30 Chinese provinces. Several latent
class regression models are applied to capture cross-regional heterogeneity and investigate the
determinants of output, income, and employment multipliers. Results indicate that the level of
economic development is positively associated with output multiplier, and it is also positively
associated with employment multiplier when the multiplier is large. Moreover, the size of
tourism economy is negatively associated with employment multiplier when the multiplier is
small. Some policy implications are presented based on empirical results.
Keywords: tourism impact, latent class regression, input-output analysis, economic multiplier(s)
Introduction
The tourism industry enjoys a well established reputation for exerting substantial multiplier
effects on local and regional economies. As such, the importance of tourism and its economic
impacts across local and regional economies is the focus of many studies, and is a topic that
continues to gain attention as scholars attempt to deepen their understanding of the implications
of tourism as a regional economic base (Archer, 1995; Croes and Vanegas, 2008; Vaughan et al.,
2000). Ryan (2003) suggests that the general economic benefits of tourism are many, ranging
from the generation of foreign exchange, tax revenue, and employment opportunities, to the
circulation and redistributing of income. As tourist dollars flow and disperse throughout a local

2
economy, affected sectors and linkages are stimulated, producing a multiplier effect that is both
internal to the local economy and spillover effects which are regional. Subsequently, the
economic contributions made by the tourism industry include various multiplier effects; all of
which are inherently spatial. Multipliers have been calibrated to measure the specific economic
importance of tourism to the local economy; these include the estimation of output multipliers,
income multipliers, employment multipliers, and government revenue multipliers (Gasparino et
al., 2009). With the advancement of modern impact analysis models, these impacts can be more
accurately measured, as well as the extent to which they capture the nature of inter-sectoral
linkages and the dynamics of regional and national income, output, and benefits. These models
include the regional economic base model (Hefner, 1990), the input-output (I-O) model (West
and Gamage, 2001), the social accounting matrix (SAM) model (Daniels et al., 2004), and the
computable general equilibrium (CGE) model (Sugiyarto et al., 2003).
In the tourism-related economic impact literature, a wide variation in the magnitude of tourism’s
economic impact is observed across different regions (Lamonica and Mattioli, 2015; Munjal,
2013; Khanal et al., 2014; Klijs et al., 2015). Archer (1995) and Heng and Low (1990)
highlighted the great economic importance of tourism to local economies, while Mbaiwa (2005)
and Stoeckl (2007) found that tourism brought very limited contributions to local economies. By
using I-O analysis, Heng and Low (1990) found that the tourism income multiplier in Singapore
was 0.98, while Lee (1996) calibrated the income multiplier of tourism in South Korea to be only
0.49. Other tourism income multiplier results include 0.65 for metropolitan Victoria, Canada
(Liu and Var, 1983); 0.80 for Hawaii (Liu, 1986); 1.67 for Ely, Minnesota (Lichty and Steinnes,
1982), 0.38 for Kyongju, South Korea (Kim and Kim, 1998); and 0.40 for Tanzania (Kweka et

3
al., 2003). To unveil the forces and factors that contribute to variability in multipliers, numerous
studies have attempted to investigated the determinants of region-based multipliers as identified
in the scholarly literature (Baaijens et al., 1998; van Leeuwen et al., 2006; van Leeuwen et al.,
2009) and the implications of comparative studies using I-O tables across different regions (Pratt,
2015) and across different countries (Robles Teigeiro and Díaz, 2014).
Though a good deal of attention has been paid to the topic of regional I-O multipliers, there is
still much more that can be gleaned from investigating micro- and meso-scale variation of
tourism-related impacts, especially by examining cross-sectional input-output samples from
regions that operate under the same or a similar national economic accounting system. This
research effort could shed light on the essential factors explaining variability in the benefits of
tourism across local economies within that system. To fill this research gap, the I-O analysis
applied in this study will appraise the economic impact of the tourism industry on the local
economy of 30 provinces in China for the year 2002. A variety of multiplier estimates of tourism
are obtained, including output, income, and employment multipliers. In addition, a latent class
regression model is utilized to identify the determinants of these tourism multipliers and the
several latent classes of province-related tourism impacts. This study contributes to the current
knowledge of tourism impacts in several ways. First, we measure and compare the multiplier
effects of the regional tourism industry within the China’s economy using highly homogeneous
regional I-O tables. This comparison is reasonable as it controls for other confounding factors
like inconsistencies in the national economic account framework, distinct aggregation criteria in
I-O table construction, varying definitions of tourism, and variability in economic data quality
and availability. Second, as suggested by several evolutionary models of tourism (Butler, 1980;

4
Pearce, 1995), we suspect that the way the tourism industry interacts with other sectors is
different depending on the development stages of the industry (Pratt, 2011); and as a result, the
mechanism determining various multipliers tends to be heterogeneous across different provinces.
Our proposed latent class modeling approach is able to capture this plausible heterogeneity and
provides a reasonable assessment of regional variability of tourism multipliers in the various
Chinese provinces, given that we are focusing on both a single accounting system and a given
point in time.
Literature Review
Economic impact analysis is an important tool for tourism planners to estimate the economic
benefits generated by tourism (Tyrrell and Johnston, 2006). To this end, several methods have
gained popularity, each with its own set of advantages and limitations. Frequently utilized
tourism economic impact models include the economic base model (Hefner, 1990), the input-
output (I-O) model (West and Gamage, 2001), the social accounting matrix (SAM) model
(Daniels et al., 2004; Polo and Valle, 2009), the computable general equilibrium (CGE) model
(Sugiyarto et al., 2003), and the tourism satellite account (TSA) model (Frechtling, 2010). Based
on various tourism economic impact models, the variation in the magnitude of tourism’s
economic impact has been observed across different regions. Wanhill (1994) underscored the
importance of economic base in shaping the size of tourism multipliers, and a larger size of
multiplier can be found in regions with a broader economic base. By reviewing past estimates of
tourism multipliers, Lee (1996) suggested that the magnitude of the multipliers depend on the
strength of inter-sectoral linkages, the degree of leakage to outside economies, and the general
size of the economy. Pratt (2011) found that tourism’s economic contribution can be explained
by import propensity of tourists’ spend, and import propensities and linkages of tourism oriented

5
sectors. In a more recent study, Pratt (2015) confirmed that a positive relationship between
tourism multiplier and tourism economy size/level of economic development based on Chinese
provincial data. Wiersma et al. (2004) highlighted a positive association between tourism output
multiplier and region’s population and a negative association between the employment
multipliers and region’s population. According to the estimates of tourism’s impact of several
Norwegian towns, Huse et al. (1998) discovered that the economic impact hinges greatly upon
local infrastructure and the attributes and age of the local tourism industry. Ryan (2003)
identified several major determinants of tourism’s economic impacts, which include the level of
economic development, the nature of tourist attractions and facilities, ownership of tourism
business, the utilization of labor and employment levels, government provision of infrastructure,
tourist type, and the linkage to other economic sectors.
By analyzing the separate I-O tables of 114 regions in five states of the U.S., Chang (2001)
found that tourism economic multipliers can be explained by a region’s population size and the
classification of regions as “rural”, “small metro”, “large metro” or “state”. Furthermore, using a
survey data from 429 different tourism enterprises in Northern Australia, Stoeckl (2007) showed
that business-level multipliers of the tourism industry are negatively correlated with the
remoteness of location. She argued that this may be explained by the fact that tourism is weakly
linked to major or heavy industry and/or commercial agriculture. Robles Teigeiro and Díaz
(2014) estimated the multipliers for hotels and restaurants using the I-O tables from 40 countries,
and highlighted the importance of population, GDP per capita, total imports (relative to GDP),
and complexity of a region’s economic structure in explaining variability in the Rasmussen
backward multipliers for hotels and restaurants.

6
To synchronize the results from previous studies and provide a systematic review of tourism’s
economic impact, meta-regression has also been used to explore possible factors influencing
tourism multipliers based on past studies (Baaijens et al., 1998; van Leeuwen et al., 2006; van
Leeuwen et al., 2009). In particular, van Leeuwen, Nijkamp, & Rietveld (2006) discovered that
the size of the tourism multiplier is directly associated with visitor numbers and tourist
expenditures, the number and characteristics of natural and cultural attractions in the area, and
the level of administrative involvement. Moreover, van Leeuwen et al. (2009) found that the size
of the economic base and total tourism demand were the predominant factors that helped explain
the variation of documented tourism multipliers.
However, one major research direction that appears to be overlooked in extant literature is the
heterogeneity of tourism multiplier determinants. Regions with unique market accessibility,
varied resource endowments, and differing stages of tourism growth may rely on a different set
of factors as multiplier determinants (Pratt, 2011; Yang and Fik, 2014). Using a latent class
regression model, we address this gap, and aim to identify the determinants of tourism
multipliers in 30 provinces in China along with different latent classes that might reveal cross-
regional heterogeneity in tourism multiplier modeling.
Research Methods and Data
There has been a long tradition in tourism studies of using I-O models to evaluate the regional
economic impact brought about by the tourism industry. In this study, I-O models are applied to
calibrate various multipliers of the tourism industry. The rationale for choosing the I-O model is
twofold. First, the I-O model has been extensively applied in tourism research and it is particular

7
suitable for the analysis of inter-sectoral relations (Cai et al., 2006; West and Gamage, 2001;
Frechtling and Horváth, 1999; Fletcher, 1989). Klijs et al. (2012) evaluated the performance of
five economic impact models of tourism and suggested that the I-O models represented a
reasonable compromise in the economic impact assessment after taking several essential criteria
into consideration. Second, due to the unavailability of social accounting matrix data, more
sophisticated models, like SAM and CGE models, require a tremendous effort in terms of data
collection and accounting when involving multiple regions. All in all, the I-O model offers a
reasonable approach to capturing inter-sectoral linkages and estimating multipliers (Isard et al.,
1998). We calculated a set of multipliers to reflect the impact of tourism on the overall economy.
The first measure was the output multiplier, which measures the degree of total
interconnectedness between tourism and other industries (Isard et al., 1998). Moreover, we used
alternative measures to investigate the economic impact of final demand by household income
and employment rather than focusing exclusively on the gross output of each industry. Hence,
we further calibrated income and employment multipliers (Miller and Blair, 1985).
Pratt (2011) investigated the economic impact of tourism at different tourism-area-life-cycle
stages, and found that this impact achieves the maximum at the onset of the stagnation stage. In
this study, a latent class model was introduced to uncover heterogeneity in tourism multiplier
determinants. The regression coefficients in this model were specified to remain the same for
observations within the same latent class and to vary across different classes. The latent class
regression model specifies the density of a dependent variable y, as a linear combination of J
different densities, which is further depicted as
11
(|,,) (|, ), 0 1, =1
JJ
jj j j j
ii i ii i i
jj
fyx f yx
βπ π β π π
= =
= ≤≤
∑∑
(1)

8
where i indexes the observation and j indexes the latent class, j = 1, …, J.
(|, )
jj
ii
f yx
β
is the
density of jth class , and
j
i
π
is the probability of being jth class. By allowing
β
to vary across
different classes, we assumed
() j
ij i i
yx
βε
′
= +
and
2
~ (0, )
ij
N
εσ
for the jth class regression. To
parameterize the model, a constant-only multinomial logit model for
j
i
π
was specified (Cameron
and Trivedi, 2009) as:
1
exp( ) , 0
exp( )
j
j
iJ
J
m
m
θ
πθ
θ
=
= =
∑
; (2)
where the Jth coefficient of θ is set to be zero for identification purposes. By incorporating
Equation 2 into Equation 1, the parameters {
β
,
σ
, and
θ
} can be obtained by the maximization
of the appropriate log-likelihood function. To empirically determine the optimal value of J – the
total number of classes in the latent class model, one may compare the various information
criteria (e.g., Akiake Information Criteria, AIC or Bayesian Information criteria, BIC) associated
with different J values obtained from maximum likelihood estimation (Clark et al., 2005;
Bhatnagar and Ghose, 2004). In general, lower values of information criteria measures are
preferred. After the estimation of the model’s parameters, the posterior probability of class
membership for each observation i can be derived using Bayes theorem.
I-O table data were obtained from the China National Statistics Bureau for 42 sectors (i.e.,
industries) in 2002 in each of 30 selected Chinese provinces (National Bureau of Statistics of
China, 2008). We isolated two of these 42 sectors as being reasonably representative of tourism:
the “original” tourism industry and the hotel and restaurant industries, collectively. The former is
defined as the industry that provides particular services to packaged and non-packaged tourists

9
and business travelers, which normally includes travel agencies and tourism attractions.
Compared to the I-O tables used in previous research, the data utilized in this study possesses
several advantages. First, a specific tourism industry has been identified as a unique and
integrated sector in the national economic account, and therefore, it inherently avoids possible
aggregation bias associated with aggregating the tourism industry from more than one of the
conventional industries in the I-O table (Briassoulis, 1991). Second, the data provides valuable
information on regional heterogeneity in terms of economic structure. Since all of the N=30 I-O
tables were calibrated under the supervision of China’s central government and based on
identical methods, the industrial linkage and multiplier measures can be directly comparable
across different provinces. Therefore, the data sets used herein are able to generate more reliable
comparison compared to the international data set (Robles Teigeiro and Díaz, 2014). However, it
should be noted that the physical labor input coefficient is not available in these I-O tables.
Nevertheless, sector employment data was obtained from the Labor Statistical Yearbook of
China (2003). For the narrowly defined manufacturing and mining sectors, we also filled in
unavailable employment data within each broad sector by assuming that the employment-to-
compensation ratio was consistent all over the nested and narrow sectors.
Dependent variables of the regression were log of output, income, and employment multipliers
as indicated in the I-O analysis. We selected independent variables that are commonly identified
in the literature; see Table 1 for variable definitions. Among them, lnGDP_per represented the
level of economic development, with a more developed economy deemed to be more capable
than less developed economies of internalizing economic spillover generated by tourism due to
having more economic linkages (i.e., because of said economy’s large size) (Lee, 1996; van

10
Leeuwen et al., 2009). Moreover, lnTour captures the size of provincial tourism economy, and it
was used as a proxy for localization economies (Majewska, 2015), and according to new
economic geography, a clustering of related sectors tends to improve productivity and facilitate
inter-sector linkage across different industries (Ryan, 2003; Pratt, 2011). Lastly, because
transport infrastructure has been recognized as a factor facilitating tourism growth (Pratt, 2011),
we used lnFlights_per, as an indicator of air transport infrastructure. The data for these
independent variables were obtained from the China Statistical Yearbook (2003) and Department
of Development and Planning CAAC (2003). Corresponding descriptive statistics are presented
in Table 1.
(Please insert Table 1 about here)
Results and Discussion
The multiplier measures obtained from I-O analysis serve as important indicators to assess the
economic contribution of tourism to the local economy, and a high tourism multiplier measure
indicates large economic benefits that tourism generates to the economy. Table 2 presents the
results of three tourism multiplier measures in 30 Chinese provinces after I-O analysis, which
indicated substantial variation across provinces. Recall that an output multiplier of tourism
measures the total value of production in the economy that is necessary to satisfy a unit of output
for the tourism industry. The total output multiplier of Beijing is estimated to be 2.402,
suggesting that additional outputs of 2.402 RMB Yuan from all sectors in the economy are
required for 1 RMB Yuan new final demand for the output of the tourism industry in Beijing.
Tianjin, Shanghai, and Inner Mongolia are found to have the largest output multiplier of tourism
in all of 30 China’s provinces, indicating that the tourism industry in these three provinces is
highly connected with other industries as a purchaser. On the other hand, Yunnan, Jiangsu, and

11
Hebei are characterized by the smallest output multiplier of tourism, showing the limited
importance of tourism as a purchaser in the local economy. According to Table 2, the income
multiplier is estimated to be the largest in Jiangxi, Guizhou, and Liaoning, indicating the
substantial economic impact of tourism on local household income. On the other hand, this
multiplier is found to the smallest in Hebei, Yunnan, and Jiangsu. Regarding the employment
multiplier, Jilin, Heilongjiang, and Inner Mongolia are ranked in top three.
(Please insert Table 2 about here)
The maps of three tourism multipliers are presented in Figure 1. It shows a spatial clustering of
values in output multiplier and income multiplier in the south, and a clustering of values in
employment multiplier in the north. An application of global Moran’s I statistics, which measure
the degree of spatial dependence or spatial autocorrelation, reveals a significant and positive
spatial dependence of employment multipliers among the provinces (Anselin, 1995). Using a
first-order contiguity spatial weighting matrix, the global Moran’s I of employment multiplier is
estimated to be 0.339 with a pseudo p-value of 0.009. The results suggest the presence of
regional based spillovers in tourism-related employment, or spatial linkages between sectors that
potentially spillover over province boundaries (Yang and Fik, 2014).
(Please insert Figure 1 about here)
Following the method used by Pratt (2015), we evaluated the bi-variate relationship between
dependent and independent variables before the regression analysis. Figure 2 presents the bi-
variate scatter plot without logarithm transformation. In general, a positive association

12
characterizes the relationship between output multiplier and independent variables, whereas a
negative relationship characterized the relationship between income/employment multiplier and
independent variables. Moreover, we find that the distribution of some variables are left-skewed,
and a logarithm transformation is necessary.
(Please insert Figure 2 about here)
A latent class regression model was further introduced to identify the factors behind this
variation and to capture heterogeneity in tourism multiplier models. We found a high level of
correlation between lnGDP_per and lnFlights_per with a Person correlation coefficient of 0.741.
Therefore, we only included two independent variables, lnGDP_per and lnTour in further
modeling efforts. Before estimating the model, we used information criteria to determine the
optimum number of latent classes to include in the model (see Table 3). Since we only have 30
observations, a maximum of two latent classes was considered. A single-class model was
selected for the output and income multiplier models, and a 2-class model was chosen for the
employment multiplier model with lnGDP_per as independent variable in one class and lnTour
in the other class.
(Please insert Table 3 about here)
Table 4 presents the estimation results of the latent regression models. In Model 1 for output
multiplier, the results show that only the estimated coefficient of lnGDP_per is statistically
significant. Its coefficient suggests that a 1% increase in GDP per capita will lead to a 0.072%
increase in output multiplier of regional tourism. This finding can be explained by the fact that a

13
more developed economy provides more diversified economic structure and is thus more likely
to absorb the leakage effect from tourism (Huse et al., 1998). A diversified economy enables the
tourism industry to find suitable local suppliers, thereby contributing to a higher level of
interconnectedness with other industries (van Leeuwen et al., 2009; Stoeckl, 2007; Robles
Teigeiro and Díaz, 2014). In Model 2, which assessed the income multiplier, neither of the
specified independent variables was found to be statistically significant. In Model 3 which
contains two latent classes related to the employment multiplier, different significant variables
were identified for different classes. For class 1, lnGDP_per was estimated to be statistically
significant and positive, therefore suggesting that the level of economic development is a major
determinant of this multiplier. For class 2, lnTour was estimated to be significant and negative.
One possible explanation for this result is that in regions with a large size of tourism economy,
the economies of scale improves the efficiency of labor allocation related to the tourism industry,
and therefore, the net impact of stimulated employment per capita would decrease. We predicted
the class membership of each province and discovered that the employment multiplier was
significantly larger in class 1 provinces. Therefore, these results indicate that when the
employment multiplier is small, a large size of tourism economy would shrink the employment
multiplier, and when the multiplier is large, a developed local economy would boost the
employment multiplier. Furthermore, we found that geographical location and political hierarchy
cannot reasonably explain the membership of latent classes.
(Please insert Table 4 about here)

14
Conclusions
This study calibrated the output, income and employment multipliers of tourism across 30
Chinese provincial regions and applied latent class regression models to understand the factors
explaining these multipliers. The estimation results suggest that the level of economic
development is positively related to output multiplier and employment multiplier when the
employment multiplier is large. Also, the size of tourism economy is negatively related to
employment multiplier. We made some theoretical contributions to the past literature. First of all,
we presented a more comprehensive tourism multiplier analysis using a cross-sectional sample of
regions to explain three types of multipliers. Second and more importantly, we highlighted the
cross-regional heterogeneity in tourism multiplier determinants, and different factors may have
different impacts for tourism multiplier across regions. Third, we found some empirical evidence
on the negative relationship between tourism economy size and employment multiplier, and this
result underscored the essential differences across different types of tourism multipliers.
This study’s results present important implications for local governments related to enlarging
tourism multipliers and improving the overall economic contribution of tourism. In general, our
results underscored the importance economic development in stimulating tourism output
multipliers; tourism injection is more rewarding for more economically developed regions than
less developed ones. Furthermore, as reflected in the findings, the “one-size-fits-all” tourism
policies that currently overlook regional heterogeneity may lead to disappointing results.
Previous methods used to categorize Chinese provinces, such as geographical location and
political hierarchy, were found to be ineffective in understanding membership according to the
present study’s latent class models. In understanding employment multipliers, the level of this
multiplier was found to be a major indicator by which to classify provinces.

15
We would like to encourage future studies to apply a similar analysis to cross-sectional samples
covering multiple regions, states, and cities. Future research along these lines could also include
a longitudinal study to trace the trajectories and dynamics of inter-industrial linkages as they
pertain to tourism and tourist-related services.
References
Anselin L. (1995) Local indicators of spatial association-LISA. Geographical Analysis 27: 93-
115.
Archer B. (1995) Importance of tourism for the economy of Bermuda. Annals of Tourism
Research 22: 918-930.
Baaijens SR, Nijkamp P and Van Montfort K. (1998) Explanatory meta-analysis for the
comparison and transfer of regional tourist Income multipliers. Regional Studies 32: 839-
849.
Bhatnagar A and Ghose S. (2004) A latent class segmentation analysis of e-shoppers. Journal of
Business Research 57: 758-767.
Briassoulis H. (1991) Methodological issues: Tourism input-output analysis. Annals of Tourism
Research 18: 485-495.
Butler R. (1980) The concept of a tourist area cycle of evolution: Implications for management
of resources. Canadian Geographer 24: 5-12.
Cai J, Leung P and Mak J. (2006) Tourism’s forward and backward linkages. Journal of Travel
Research 45: 36-52.

16
Cameron AC and Trivedi PK. (2009) Microeconometrics using Stata, College Station, TX: Stata
Press.
Chang WH. (2001) Variations in multipliers and related economic ratios for recreation and
tourism impact analysis. Dept. of Park, Recreation and Tourism Resources. Michigan
State University.
Clark A, Etilé F, Postel-Vinay F, et al. (2005) Heterogeneity in reported well-being: Evidence
from twelve European countries. The Economic Journal 115: C118-C132.
Croes R and Vanegas M. (2008) Cointegration and causality between tourism and poverty
reduction. Journal of Travel Research 47: 94-103.
Daniels MJ, Norman WC and Henry MS. (2004) Estimating income effects of a sport tourism
event. Annals of Tourism Research 31: 180-199.
Department of Development and Planning CAAC. (2003) Statistical Data of Civic Aviation in
China, 2003, Beijing: China Civil Aviation Press.
Fletcher JE. (1989) Input-output analysis and tourism impact studies. Annals of Tourism
Research 16: 514-529.
Frechtling DC. (2010) The tourism satellite account: A primer. Annals of Tourism Research 37:
136-153.
Frechtling DC and Horváth E. (1999) Estimating the multiplier effects of tourism expenditures
on a local economy through a regional input-output model. Journal of Travel Research
37: 324-332.
Gasparino U, Bellini E, Del Corpo B, et al. (2009) Measuring impact of tourism on urban
economies: A review of literature. International Journal of Leisure and Tourism
Marketing 1: 103-130.

17
Hefner FL. (1990) Using economic models to measure the impact of sports on local economies.
Journal of Sport & Social Issues 14: 1-13.
Heng TM and Low L. (1990) Economic impact of tourism in Singapore. Annals of Tourism
Research 17: 246-269.
Huse M, Gustavsen T and Almedal S. (1998) Tourism impact comparisons among Norwegian
towns. Annals of Tourism Research 25: 721-738.
Isard W, Azis IJ, Drennan MP, et al. (1998) Methods of Interregional and Regional Analysis,
Aldershot, Hants, England: Ashgate.
Khanal BR, Gan C and Becken S. (2014) Tourism inter-industry linkages in the Lao PDR
economy: An input—output analysis. Tourism Economics 20: 171-194.
Kim S-H and Kim K-H. (1998) Impact of tourism on local economies: An income multiplier
analysis. Asia Pacific Journal of Tourism Research 2: 49-56.
Klijs J, Heijman W, Maris DK, et al. (2012) Criteria for comparing economic impact models of
tourism. Tourism Economics 18: 1175-1202.
Klijs J, Peerlings J and Heijman W. (2015) Usefulness of non-linear input—output models for
economic impact analyses in tourism and recreation. Tourism Economics 21: 931-956.
Kweka J, Morrissey O and Blake A. (2003) The economic potential of tourism in Tanzania.
Journal of International Development 15: 335-351.
Lamonica GR and Mattioli E. (2015) Research Note: The Impact of the Tourism Industry on the
World's Largest Economies —An Input—Output Analysis. Tourism Economics 21: 419-
426.
Lee C-K. (1996) Input-output analysis and income distribution patterns of the tourism industry in
South Korea. Asia Pacific Journal of Tourism Research 1: 35-49.

18
Lichty RW and Steinnes DN. (1982) Measuring the impact of tourism on a small community.
Growth and Change 13: 36-39.
Liu J and Var T. (1983) The economic impact of tourism in Metropolitan Victoria, B.C. Journal
of Travel Research 22: 8-15.
Liu JC. (1986) Relative economic contributions of visitor groups in Hawaii. Journal of Travel
Research 25: 2-9.
Majewska J. (2015) Inter-regional agglomeration effects in tourism in Poland. Tourism
Geographies 17: 408-436.
Mbaiwa JE. (2005) Enclave tourism and its socio-economic impacts in the Okavango Delta,
Botswana. Tourism Management 26: 157-172.
Miller RE and Blair PD. (1985) Input-Output Analysis: Foundations and Extensions, Englewood
Cliffs, N.J: Prentice-Hall.
Munjal P. (2013) Measuring the economic impact of the tourism industry in India using the
Tourism Satellite Account and input—output analysis. Tourism Economics 19: 1345-
1359.
National Bureau of Statistics of China. (2008) China's Regional Input-Output Table, Beijing:
China Statistics Press.
Pearce DG. (1995) Tourism Today: A Geographical Analysis, Harlow, Essex, England:
Longman Scientific & Technical.
Polo C and Valle E. (2009) Estimating tourism impacts using input–output and SAM models in
the Balearic islands. In: Matias Á, Nijkamp P and Sarmento M (eds) Advances in
Tourism Economics. Physica-Verlag HD, 121-143.

19
Pratt S. (2011) Economic linkages and impacts across the TALC. Annals of Tourism Research
38: 630-650.
Pratt S. (2015) Potential economic contribution of regional tourism development in China: A
comparative analysis. International Journal of Tourism Research 17: 303-312.
Robles Teigeiro L and Díaz B. (2014) Estimation of multipliers for the activity of hotels and
restaurants. Tourism Management 40: 27-34.
Ryan C. (2003) Recreational Tourism: Demand and Impacts, Buffalo, N.Y. : Multilingual
Matters.
Stoeckl N. (2007) Using surveys of business expenditure to draw inferences about the size of
regional multipliers: A case-study of tourism in northern Australia. Regional Studies 41:
917-931.
Sugiyarto G, Blake A and Sinclair MT. (2003) Tourism and globalization: Economic Impact in
Indonesia. Annals of Tourism Research 30: 683-701.
Tyrrell TJ and Johnston RJ. (2006) The economic impacts of tourism: A special issue. Journal of
Travel Research 45: 3-7.
van Leeuwen ES, Nijkamp P and Rietveld P. (2006) Economic impacts of tourism: A meta-
analytic comparison of regional output multipliers. In: M G and P N (eds) Tourism and
Regional Development: New pathways. Aldershot: Ashgate, 115-132.
van Leeuwen ES, Nijkamp P and Rietveld P. (2009) A meta-analytic comparison of regional
output multipliers at different spatial levels: Economic impacts of tourism. In: Matias Á,
Nijkamp P and Sarmento M (eds) Advances in Tourism Economics: New Developments.
Physica-Verlag HD, 13-33.

20
Vaughan DR, Farr H and Slee RW. (2000) Estimating and interpreting the local economic
benefits of visitor spending: An explanation. Leisure Studies 19: 95-118.
Wanhill S. (1994) The measurement of tourist income multipliers. Tourism Management 15:
281-283.
West G and Gamage A. (2001) Macro effects of tourism in Victoria, Australia: A nonlinear
input-output approach. Journal of Travel Research 40: 101-109.
Wiersma J, Morris D and Robertson R. (2004) Variations in Economic Multipliers of the
Tourism Sector in New Hampshire. Northeastern Recreation Research Symposium.
Bolton Landing, New York.
Yang Y and Fik T. (2014) Spatial effects in regional tourism growth. Annals of Tourism
Research 46: 144-162.

1
Table 1. Descriptive statistics of independent variables
Variable
Definition
Mean
Std.
Dev.
Min
Max
ln
GDP_per
Log of GDP per capita (in 1,000
RMB)
9.068
0.570
8.056
10.612
ln
Tour
Log of inbound tourist arrivals
13.275
1.374
8.703
16.451
ln
Flights_per
Log of air flights in civic
airports divided by population
(in 10,000)
2.575
1.089
-0.244
5.148

2
Table 2. Tourism economic multipliers in different provinces
Province
Output
multiplier
Rank
Income
multiplier
Rank
Employment
multiplier
Rank
Beijing
2.402
16
1.974
27
1.495
30
Tianjin
2.900
1
2.286
21
2.731
18
Hebei
2.074
30
1.629
30
2.341
21
Shanxi
2.185
27
3.351
6
3.005
15
Inner
Mongolia
2.767
3
2.528
17
3.948
3
Liaoning
2.378
17
4.050
3
3.616
6
Jilin
2.501
9
2.966
12
5.035
1
Heilongjiang
2.475
12
3.024
11
4.856
2
Shanghai
2.836
2
2.175
23
2.324
22
Jiangsu
2.128
29
1.914
28
1.896
27
Zhejiang
2.499
11
3.199
7
1.817
28
Anhui
2.636
8
3.039
9
3.719
5
Fujian
2.660
6
3.031
10
2.645
20
Jiangxi
2.427
14
4.266
1
3.497
8
Shandong
2.668
5
2.923
14
3.443
9
Henan
2.374
18
2.148
24
1.923
26
Hubei
2.296
22
2.272
22
2.661
19
Hunan
2.500
10
2.962
13
2.315
23
Guangdong
2.721
4
3.161
8
2.009
25
Guangxi
2.421
15
3.630
5
3.025
14
Hainan
2.276
23
2.693
15
3.141
13
Chongqing
2.350
21
2.525
18
2.948
16
Sichuan
2.214
25
2.584
16
3.606
7
Guizhou
2.469
13
4.220
2
3.293
11
Yunnan
2.140
28
1.794
29
2.130
24
Shaanxi
2.360
20
2.457
20
1.574
29
Gansu
2.366
19
2.076
25
3.301
10
Qinghai
2.191
26
1.980
26
3.217
12
Ningxia
2.644
7
3.894
4
3.872
4
Xinjiang
2.269
24
2.470
19
2.807
17

3
Table 3. Information criteria values of different latent class specifications
1-class
2-class
2-class
2-class
2-class
Independent variables lnGDP_per
and lnTour
lnGDP_per
and lnTour
lnGDP_per
or lnTour
lnGDP_per lnTour
Output multiplier
AIC
-59.331
-54.429
-50.249
-55.699
-59.726
BIC
-53.726
-41.818
-40.441
-45.890
-49.918
Income multiplier
AIC
9.149
16.028
12.239
14.066
14.281
BIC
14.754
28.639
22.048
23.875
24.090
Employment multiplier
AIC
11.931
15.448
2.689
16.065
17.834
BIC
17.536
28.059
11.096
25.874
27.642

4
Table 4. Estimation results of latent class regression models
Model 1
Model 2
Model 3
Output
multiplier
Income
multiplier
Employment multiplier
Class 1
Class 2
lnGDP_per
0.0720**
-0.0935
0.584***
(0.029)
(0.091)
(0.004)
lnTour
-0.00847
-0.0128
-0.115***
(0.014)
(0.046)
(0.025)
constant
0.347
2.006***
-3.816***
2.528***
(0.227)
(0.691)
(0.039)
(0.328)
σ
j
0.00621***
0.0608***
0.00000509**
0.0577***
(0.001)
(0.013)
(0.000)
(0.015)
P(j|i)
0.094
0.906
log-
likelihood
33.666
-0.574
4.656
Notes: * indicates p<0.10, ** indicates p<0.05, *** indicates p<0.01. Robust standard errors are
presented
in parentheses.

1
Figure 1. Spatial distribution of tourism multipliers over Chinese provinces
Output multiplier
Income multiplier
Employment multiplier

2
r = 0.451
22.22.42.6 2.8 3
010000 20000 30000 40000
GDP_per
r = 0.272
22.22.42.6 2.8 3
05.0e+06 1.0e+07 1.5e+07
Tour
r = 0.177
22.22.42.6 2.8 3
050 100 150 200
Fights_per
r = -0.252
1 2 3 4
010000 20000 30000 40000
GDP_per
Income Multiplier
r = 0.014
12 3 4
05.0e+06 1.0e+07 1.5e+07
Tour
r = -0.251
1 2 3 4
050 100 150 200
Fights_per
r = -0.343
12345
010000 20000 30000 40000
GDP_per
Employment Multiplier
r = -0.359
1 2 345
05.0e+06 1.0e+07 1.5e+07
Tour
r = -0.367
12345
050 100 150 200
Fights_per
Output Multiplier
Figure 2. Bivariate analysis dependent and independent variables.