Does the Demographic Structure Affect Outbound Tourism Demand? A Panel Smooth Transition Regression Approach

Abstract

This study aims to examine the relationship between the demographic structure and outbound tourism demand using panel data from 72 countries and regions during the 2000-2014 period. A panel smooth transition regression (PSTR) model is employed to prove that there is a nonlinear relationship between demographic factors and outbound tourism demand and that the demographic effects can vary under different income regimes. The empirical results indicate that the impacts of demographic factors on outbound tourism demand change significantly when income constraints are relaxed. Based on this premise, we can better understand the travel characteristics of different groups, which are subdivided according to different demographic factors.

Keywords

demographic structure outbound tourism tourism demand PSTR model

Introduction

Demographic changes, particularly population aging, are no longer future development trends; rather, they started a long time ago (Grimm et al. 2009). The baby boom generation (those born in roughly the two decades following World War II) has had and will continue to have important effects on the economy and society in many developed countries, such as the United States and Japan (Poterba 2001). Meanwhile, economic takeoffs in developing countries, such as China and India, have substantially benefited from changes in their demographic structures during the demographic transition (Wei and Hao 2010). In this context, demographic change is closely related to economic development, which has received considerable coverage in the literature (Miles 1999; Feyrer 2007). From a tourism perspective, demography is identified as a key driver of future consumer demand (Glover and Prideaux 2009) and provides both opportunities and threats (Yeoman, Schanzel, and Smith 2013). Considerable impacts on tourism are expected in terms of changing travel intensity, travel behavior, and tourism activities (Steiger 2012). Consequently, concerns about the relationship between demographic characteristics and outbound tourism demand are of great significance, with international tourism playing an increasingly important role in national and global economies.

The demographic structure can affect outbound tourism demand through various channels. For example, population aging is emerging as a major and predictable demographic trend worldwide (Bloom, Canning, and Fink 2010), and it is likely to affect the future choice of tourism activities and destinations (Glover and Prideaux 2009). As different generations reach milestones in their consumption and life cycle, the demand for tourism goods and travel experiences will change (MacKay 1997). Since tourism development processes and tourism-related activities are constructed based on gendered societies (Kinnaird and Hall 1996), women and men participate in and experience tourism differently (Uysal, McGehee, and Loker-Murphy 1996; Pritchard et al. 2007). There are significant gender differences in travel motivation, with male tourists preferring more entertainment and activities and female tourists having a greater relaxation and escape motivation (Andreu et al. 2006). In addition, the population distribution is increasingly urbanized, especially in developing countries. People located in urban and developed areas have more flexibility to travel than individuals living in rural and developing territories (Bernini and Cracolici 2015), and there are also differences in tourism demand. Among urban dwellers, the experience of urban life may generate a demand for rural tourism experiences, and vice versa (WTO and ETC 2010). Moreover, a higher level of education broadens an individual’s access to information and knowledge, and it also indirectly reflects income constraints (Bernini and Cracolici 2015). Education may affect travel motivation and the destination choice (Sangpikul 2010). In addition, there are many other important demographic factors that may have an impact on outbound tourism demand, such as life expectancy, fertility, and family size.

Although demography is clearly relevant to outbound tourism demand, it has not received sufficient attention from tourism scholars. Little is known about how the demographic structure affects outbound tourism demand since scarce research has examined the extensive channels of the influence of demography. Limited evidence that demographic factors shape outbound tourism demand is provided by studies of specific population groups, such as elderly individuals (see Fleischer and Pizam 2002; Lu et al. 2016), young generations (see X. Li, Li, and Hudson 2013) and women (see Figueroa-Domecq et al. 2015), and single-country case studies that discuss only some demographic variables (see Collins and Tisdell 2002; Szromek, Januszewska, and Romaniuk 2012).

This study aims to conduct an empirical analysis on the potential link between the demographic structure and outbound tourism demand by applying a panel smooth transition regression (PSTR) model as a new econometric technique. The main contributions are as follows: (1) This study presents new empirical evidence on the relationship between the demographic structure and outbound tourism demand in a cross-country context. (2) It divides demographic factors into two major dimensions, the natural structure and social structure, based on biological characteristics and social attributes and analyzes the heterogeneous impact of different demographic factors (i.e., the natural structure and social structure) on outbound tourism demand. (3) It employs the PSTR model proposed by A. González, Teräsvirta, and van Dijk (2005) to examine the possible relationship between demographic factors and outbound tourism demand while considering the moderating effects of income.

The rest of the article is organized as follows. The next section provides a review of the related literature to identify the relevant theoretical background for this study. The third section demonstrates the methodology and the main variables introduced in this study, which is followed by a presentation of and robustness tests of the model estimation results in the fourth section. The fifth section presents a discussion of the findings and their theoretical significance. The sixth section offers some conclusions and corresponding policy implications.

Literature Review

Demographic Factors and Outbound Tourism Demand

The demographic structure describes the composition of human populations from various aspects (S. Li and Zhou 2019). In general, the demographic structure can be divided into two major dimensions: the natural structure and the social structure. The natural structure denotes the distribution of human populations based on their biological characteristics, such as age and gender, whereas the social structure refers to the distribution of human populations according to their social attributes, such as education, urbanization and family (Liddle 2014).

From the perspective of the demographic natural structure, age and cohort have been identified as the main determinants of outbound tourism demand (You and O’Leary 2000; Mak, Carlile, and Dai 2005). Specifically, membership in a specific generation or age group influences people’s consumption patterns. Therefore, there is reason to believe that people’s tourism demand generally changes as their corresponding consumption and life cycle stages. Empirical evidence shows significant differences in the propensity to travel abroad, the length of stay and the level of expenditure overseas among different generations (X. Li, Li, and Hudson 2013) and age groups (Riker 2014; Bernini and Cracolici 2015). Another factor that has a major impact on outbound tourism demand is gender (Collins and Tisdell 2002). Gender differences related to travel and tourism clearly exist because human society is gendered (Kinnaird and Hall 1994). Previous studies have discussed various gender differences mainly with regard to travel motivation (Meng and Uysal 2008), travel patterns (Firestone and Shelton 1994), preferences for travel experiences (McCleary, Weaver, and Lan 1994; Collins and Tisdell 2002), and tourist decision-making processes (Mottiar and Quinn 2004). Eugenio-Martin and Campos-Soria (2011) proved that gender factors have a significant impact on the demand for outbound tourism. Sakai, Brown, and Mak (2000) also proposed that it is beneficial to analyze outbound tourism demand for men and women separately because they could have different reasons for travel and their travel preferences are different.

From the perspective of the demographic social structure, educational attainment also has an important impact on outbound tourism demand (Bernini and Cracolici 2015). People with a higher level of education are more likely to be motivated to participate in tourism activities (Nicolau and Más 2005). Education allows better access to information and knowledge (Hong, Morrison, and Cai 1996), and people with higher educational levels generally have better positions and higher incomes to meet their high-level tourism demand and to bear higher prices of tourism goods and services (Bernini and Cracolici 2015). Zimmer, Brayley, and Searle (1995) also proved that education influences travelers when choosing destination that are nearby or farther away, with travelers who are better educated being more likely to travel farther from home. Furthermore, the role played by population location has also been emphasized (WTO and ETC 2010). Urbanization is the main characterization of population location, which reflects the urban–rural structure of human populations. People located in urban and developed areas have more flexibility to travel than individuals living in rural and developing territories (Eugenio-Martin and Campos-Soria 2011; Bernini and Cracolici 2015). Urban residents and rural residents may have different outbound tourism demands because of different life experiences, and urban residents’ experience with urban life may create a demand for rural tourism experiences (WTO and ETC 2010).

Demographic Factors and Outbound Tourism Demand: The Moderating Effect of Income

Unsurprisingly, because international tourism is generally regarded as a luxury product (Bond and Ladman 1972), income and price are identified as the core determinants of international tourism demand (Crouch 1994; Lim 1999). Numerous studies have proven that economic factors account for much of the variation in the number of international tourists and expenditures (see P. González and Moral 1995; Lim 1997; Lin, Liu, and Song 2015). More precisely, income usually positively affects international tourism demand, with an average elasticity between 1.86 and 2.53, while price often negatively affects international tourism demand, with an average elasticity between −1.281 and −0.63 (Crouch 1996; Peng et al. 2015). Consequently, a consideration of the role of economic factors in the relationship between the demographic structure and outbound tourism demand is inevitable.

To date, scarce research has reported this possible moderating effect of income. Limited research has simply profiled travelers by income level and investigated the differences in travel behaviors and travel preferences between groups. Jang et al. (2004) identified the characteristics of Japanese pleasure travelers to the United States by income level and found that there are different expenditure patterns in different income groups. Similarly, Nicolau and Más (2005) and Kattiyapornpong and Miller (2011) explain the difference in the probability of going on holiday and the destination selection preference between different income groups. Recently, Jiang, Liu, and Song (2017) compared the influencing factors of outbound tourism demand in developing economies and developed economies. It was found that the age structure and educational level of developed and developing economies have great differences in the impact on outbound tourism demand and that developed economies are more affected.

The existing research has provided us with a rough theoretical framework regarding how demographic factors affect outbound tourism demand. Nevertheless, empirical evidence from specific population groups and single-country cases makes a limited contribution to interpreting the relationship between the demographic structure and outbound tourism demand. Because of the different economic and social development backgrounds of different countries, the stages and characteristics of outbound tourism development are not the same. Therefore, the same factors may have different effects in different countries (Qiu and Zhang 1995). In addition, profiling travelers by income level is a straightforward approach but does not accurately identify the moderating effect of income. Since the threshold value of income is given a priori, it might lead to biased estimations.

To fill this research gap, we introduce a PSTR model to identify the moderating effect of income and to examine the possible relationship between demographic factors and outbound tourism demand using data from 72 countries over a 15-year period (2000-2014). The PSTR model, which was expanded from the panel threshold regression (PTR) model proposed by Hansen (1999), uses a continuous transition function to replace the discrete indicative function in the model. Thus, it can better grasp the cross-sectional heterogeneity of panel data and allow the model parameters to realize a continuous and smooth nonlinear transformation with changes in the transition variables to closely reflect the economic reality (A. González, Teräsvirta, and van Dijk 2005). As a result of capturing nonlinearity and regime-switching effects in this way, the PSTR model is an attractive approach and is widely used in the literature (Duarte, Pinilla, and Serrano 2013). Then, we further investigate the heterogeneous impacts of the demographic natural structure and social structure on the propensity to travel abroad and tourism expenditure.

Methodology

Variable Selection and Data Sources

The measurement of outbound tourism demand usually considers two aspects: whether to participate and how much to spend. In general, the number of outbound tourists and travel costs are regarded as appropriate dependent variables for outbound tourism demand (Lim 1997; Song et al. 2010). However, these aggregate indicators cannot be compared in different regions because of the different population bases. Consequently, we select relative indicators to characterize outbound tourism demand, namely, the outbound tourism rate (OTR) and outbound tourism expenditure rate (OTER) (see Dai and Sun 2014; Gholipour, Tajaddini, and Al-Mulali 2014; Jiang, Liu, and Song 2017). The variables are defined as follows: OTR = the number of outbound tourists /total population ×100% and OTER = outbound tourism expenditure /total population. The number of outbound tourists and the expenditure data originate from the World Tourism Organization, the Yearbook of Tourism Statistics, the Compendium of Tourism Statistics, and data files. The population data come from United Nations Population Division and Census reports and other statistical publications of national statistical offices. The values of total population shown are midyear estimates. The number of outbound tourists refers to the number of departures that people make from their country of usual residence to any other country for any purpose other than a remunerated activity in the country visited. Outbound tourism expenditures are expenditures of international outbound visitors in other countries, including payments to foreign carriers for international transport (current price in US dollars).

The demographic structure is usually measured by the proportion of a specific group to the total population or the ratio of a specific group to another group (Li and Zhou 2019). Based on the demographic structure definition given above, we systematically sorted the variables from the two dimensions of the natural structure and the social structure. Starting from the inherent relationship between the demographic structure and outbound tourism development, a theoretical analysis and a screening of the variables are performed by fully considering the principle of availability and relevance. Five demographic structure variables were selected as explanatory variables, that is, the age structure (the older cohort and the cohort of individuals aged 15-64 years), gender structure, education structure, and urbanization structure. First, the proportion of the population aged 15-64 (PP) and the elderly dependency ratio (EDR) are used to measure the age structure of the population. People aged 15-64 years dominate the outbound tourism market (Dai and Sun 2014), and we cannot ignore this important group. EDR reflects the aging process of a country or region, and the importance of the elderly in the tourism market has changed significantly compared with past years. In the near future, the number of elderly individuals with a higher level of education and physical health will increase, and they may have more chances to travel abroad (Christensen et al. 2009). Older cohorts are also a major source of health and medical tourism (Yeoman, Schanzel, and Smith 2013). Second, the gender structure is measured by the proportion of the female population (PFP). There are differences in travel behavior between men and women (Frew and Shaw 1999), mainly because tourism development processes and tourism-related activities are constructed based on gendered societies (Kinnaird and Hall 1996). The importance of women as tourism consumers warrants attention in the context of the increasing promotion of gender equality and women’s empowerment. Currently, most women are economically independent because they work, and they no longer need to rely on men (fathers or husbands) to take care of them and their offspring (Yeoman et al. 2011). This independence allows them to decide whether to spend their income on travel or not. In addition, people’s attitudes towards marriage and singleness have been changing for decades, and divorce rates have risen. As a result, there has been an increase in single women and women without family constraints who have more free time, constitute a proportion of the tourism market, and tend to allocate more income to tourism (Yeoman et al. 2011). Therefore, a focus on the proportion of the female population helps to identify the relative importance of female travelers to the global outbound travel market. Third, the proportion of the urban population (PUP) is used to measure the urbanization structure, that is, the urban–rural structure. Urban residents are very different from rural residents in terms of their tourism motivation, destination choices, and tourism consumption. In the process of urbanization, the mutual influence between urban and rural residents’ concepts and patterns of consumption may have a positive impact on the development of the outbound tourism market. Fourth, the higher education ratio (HER), which is presented by the percentage of the college (ISCED 5 and 6) enrollment rate, is used to evaluate the education structure. The level of education is the main determinant of the decision to participate in tourism, and people with higher educational levels are more likely to afford higher outbound travel costs and enjoy the trip (Bernini and Cracolici 2015). The demographic structure data come from United Nations Population Division’s World Population Prospects, United Nations Population Division’s World Urbanization Prospects, UNESCO Institute for Statistics.

Considering that the demographic structure changes very slowly and that a high degree of dispersion appears only across a long period of time, more samples should be included to make the research findings present a universal value. Initially, samples from 213 countries and regions in the world were collected. Then, some samples were deleted according to the principle of data consistency and availability. Finally, panel data from 72 countries and regions during the 2000–2014 period were used as research samples in this paper. According to the classification standards of the World Bank’s World Development Indicators (WDIs), the samples include 24 high-income OECD members, 9 high-income non-OECD members, 21 medium- and high-income areas, 16 medium- and low-income areas, and 2 low-income areas.¹ The selected samples cover the main outbound tourist source areas in the world. The number of outbound tourists and the amount of outbound tourism expenditure in these areas account for 83.8% and 88.1%² of the global total number of outbound tourists and the global total amount of outbound tourism expenditure, respectively, guaranteeing the representativeness and typicality of the samples. The definitions and descriptive statistics for all variables are shown in Table 1. To overcome any possible heteroscedasticity, all variables are logarithmically processed, except for the GDP growth rate (RGDP), inflation rate (INF), and savings rate (SR), which contain negative values.

Table 1.

Definitions and Descriptive Statistics for All Variables.

Variables	Variable Name	Definition	Mean	SD	Min	Max
Dependent variable
LnOTR	Outbound tourism rate (OTR)	OTR = number of outbound tourists / total population × 100%	2.9469	1.4915	−1.2133	7.0919
LnOTER	Outbound tourism expenditure rate (OTER)	OTER = outbound tourism expenditure / total population	4.9621	1.5492	0.8190	8.4070
Explanatory variable
LnPP	Proportion of population aged 15–64 years (PP)	Population aged 15–64 years (% of total population)	4.1833	0.0644	3.9721	4.3199
LnEDR	Elderly dependency ratio (EDR)	Population aged ≥65 years (% of working-age population)	2.6342	0.4917	1.6931	3.7352
LnPFP	Proportion of female population (PFP)	Female population (% of total population)	3.9276	0.0239	3.8694	3.9947
LnPUP	Proportion of urban population (PUP)	Urban population (% of total population)	4.1388	0.3721	2.5976	4.6052
LnHER	Higher education ratio (HER)	College (ISCED 5 and 6) enrollment rate (%)^a	3.6862	0.6685	0.8673	4.7514
Transition variable
LnGNI	The gross national income (GNI) per capita	GNI per capita, Atlas method (current USD)^b	8.7945	1.3795	5.4381	11.0898
Control variable
LnPD	Population density (PD)	Population per sq. km of land area	4.3699	1.4193	0.9135	8.9537
RGDP	GDP growth rate (RGDP)	Annual percentage growth rate of GDP (%)	3.7524	3.9529	−14.8142	34.5000
SR	Savings rate (SR)	Gross domestic savings (% of GDP)^c	21.7726	11.7732	−16.8146	61.1423
LnEL	Employment level (EL)	The proportion of the population aged 15 or above who engage in economic activities (%)	4.1100	0.1368	3.6889	4.4532
INF	Inflation rate (INF)	The annual percentage change of the consumer price index (%)	5.6032	8.8720	−4.4799	168.6200
LnOPEN	Degree of openness (OPEN)	The proportion of total import and export trade in GDP (% of GDP)	4.1029	0.5402	2.8447	6.0392
LnEXCR	Exchange rate of currency (EXCR)	Official exchange rate (local currency units per USD)	2.1941	2.5157	−2.4341	9.3814

College (ISCED 5 and 6) enrollment rate (%) was calculated by the percentage of college students (ISCED 5 and 6) in the school-age population (five years after secondary school).

GNI per capita is the gross national income, converted to US dollars using the World Bank Atlas method, divided by the midyear population.

Gross domestic savings are calculated as GDP less final consumption expenditure (total consumption).

Model specification

We employed a PSTR model with fixed effects, developed by A. González, Teräsvirta, and van Dijk (2005) and Fouquau, Hurlin, and Rabaud (2008), to model the possible nonlinear relationships between demographic factors and outbound tourism demand. Gross national income (GNI) per capita, a measure of the spending capacity, is used as a transition variable to probe the impacts of demographic structure factors on outbound tourism demand in different ranges of GNI per capita. The basic two-regime PSTR model with fixed effects is defined as follows:

Y_{i t} = μ_{i} + θ X_{i t} + β_{0} P S_{i t} + β_{1} P S_{i t} g (q_{i t}; γ, c) + ε_{i t}

(1)

For $i = 1, \dots, N$ , and $t = 1, \dots, T$ , where $N$ and $T$ denote the number of individuals and the size of the sample period in the panel, respectively. $Y_{i t}$ is the dependent variable and refers to the outbound tourism rate and outbound tourism expenditure rate. $P S_{i t}$ is the core explanatory variable and refers to the demographic structure factors. $q_{i t}$ is the transition variable and denotes GNI per capita. μ _i represents the fixed individual effect, and $ε_{i t}$ is the error term. $X_{i t}$ is a set of control variables that shape outbound tourism demand. The following control variables are selected in accordance with previously conducted studies by taking into account the principle of data availability and completeness and any possible problems of multicollinearity: (1) population density (PD), reflecting the population residing in a unit area (see Nicolau and Más 2005); (2) the GDP growth rate (RGDP), is a proxy for economic growth and reflecting how fast the economy is growing (see Tugcu 2014); (3) the savings rate (SR), reflecting the proportion of gross domestic savings in GDP (% of GDP) (see Y. S. Wang 2014); (4) the employment level (EL), measured by the labor force participation rate, reflecting the proportion of the population aged 15 or older engaging in economic activities (see Dai and Sun 2014); (5) the inflation rate (INF), reflecting the annual inflation rate, measured by the consumer price index (CPI), which reflects the price fluctuation level (see Prideaux 2005); (6) the degree of openness (OPEN), reflecting the proportion of total import and export trade in GDP (% of GDP)³ (see Kulendran and Witt 2003); and (7) the exchange rate of currency (EXCR), reflecting the exchange rate of a local currency unit against the USD (see De Vita 2014). Time fixed effects also controlled. All of the above data come from the World Bank Database.

The transition function $g (q_{i t}; γ, c)$ is a continuous function of the transition variable $q_{i t}$ (e.g., GNI per capita) and is normalized to be bounded between 0 and 1, with these extreme values being associated with regression coefficients $β_{0}$ and $β_{0} + β_{1}$ . Its logistic function is set as follows:

\begin{array}{l} g (q_{i t}; γ, c) = {[1 + \exp (- γ \prod_{j = 1}^{m} (q_{i t} - c_{j}))]}^{- 1}, \\ γ > 0, c_{1} < c_{2} < \dots \leq c_{m}, \end{array}

(2)

where $c = {(c_{1}, \dots, c_{m})}^{'}$ is an m-dimensional vector of the location parameters (also known as the threshold level); it determines where the transition occurs. $m$ denotes the number of location parameters. The slope parameter $γ$ (also called a slope coefficient) determines the transition speed or the smoothness of the transitions.

A. González, Teräsvirta, and van Dijk (2005) suggested that it is usually sufficient to consider $m = 1$ or $m = 2$ empirically, as these values allow for commonly encountered types of variation in the parameters. For $m = 1$ , the PSTR model has two extreme regimes. When the transition function $g (q_{i t}; γ, c) = 0$ , the corresponding model is called a low regime; that is, the effect of the impact is in a low impact state. When $g (q_{i t}; γ, c) = 1$ , the model is called a high regime; that is, the effect of the impact is in a high impact state. Consequently, as $q_{i t}$ increases, the estimation coefficients of the model are monotonically converted from $β_{0}$ to $β_{0} + β_{1}$ , where the change is centered on $c_{1}$ . For $m = 2$ , the transition function has a minimum and symmetric value of $(c_{1} + c_{2}) / 2$ and obtains the value of 1 at both low and high values of $q_{i t}$ .

Based on the above setting, the coefficient of the influence of demographic structure factors on outbound tourism demand can be expressed as follows:

e_{i t} = \frac{\partial Y_{i t}}{\partial P S_{i t}} = β_{0} + β_{1} g (G N I_{i t}; γ, c); \forall i, \forall t

(3)

Considering $0 \leq g (G N I_{i t}; γ, c) \leq 1$ , $e_{i t}$ is actually a weighted average of parameters $β_{0}$ and $β_{1}$ . If $β_{1}$ is positive (negative), then it indicates that the effect of $P S_{i t}$ on $Y_{i t}$ increases (decreases) as the transition variable $G N I_{i t}$ increases.

Test methods

Data stationarity test

Before estimation, time-series variables must be confirmed to be stationary (Duarte, Pinilla, and Serrano 2013). To this end, we utilize the homogenous-root Levin-Lin-Chu (LLC) unit root test and the heterogeneous-root Fisher-augmented Dickey-Fuller (ADF) unit root test for panel data. If both tests reject the null hypothesis that there is a unit root, then the variable is stationary; otherwise, it is not. As shown in Table 2, the null hypothesis of nonstationarity is rejected at the 1% significance level. This result indicates that all variables are stationary and can thus be incorporated into the model.

Table 2.

Stationarity (Unit Root) Test for Panel Data.

	OTR	OTER	PP	EDR	PFP	PUP	HER	GNI
	(c, t)	(c, t)	(c, t)	(c, t)	(c, t)	(c, t)	(c, t)	(c, t)
LLC	−9.555***	−8.918***	−8.699***	−14.942***	−35.036***	−2.722***	−8.061***	−11.612***
Fisher-ADF	246.369***	192.363***	255.525***	645.029***	2576.237***	659.944***	236.559***	307.784***
	PD	EL	RGDP	SR	INF	OPEN	EXCR
	(c, t)	(c, t)	(c, t)	(c, t)	(c, t)	(c, t)	(c, t)
LLC	−9.321***	−19.935***	−11.965***	−8.815***	−11.805***	−13.347***	−27.822***
Fisher-ADF	723.293***	278.242***	311.864***	230.409***	455.343***	287.105***	292.595***

Note: (c, t) indicates that the regression equation includes both constant terms and trend terms. The null hypothesis for the LLC and Fisher-ADF tests is the presence of unit roots.

***

Significance at 1%.

Linearity test and remaining nonlinearity test

A linearity test of the PSTR model must be completed before estimation to judge whether the regime-switching effect is significant. This test is used to prove whether the null hypothesis of linearity $H_{0} : γ = 0$ is tenable. Referring to the work of A. González, Teräsvirta, and van Dijk (2005), we should replace the transition function $g (q_{i t}; γ, c)$ with its first-order Taylor expansion around $γ = 0$ and perform the linearity test through the following auxiliary regression:

Y_{i t} = μ_{i} + θ X_{i t} + β_{0} P S_{i t} + β_{1} P S_{i t} q_{i t} + \dots + β_{m} P S_{i t} q_{i t}^{m} + ε_{i t}

(4)

More precisely, testing $H_{0} : γ = 0$ is equivalent to testing the null hypothesis “ $H_{0} : β_{0} = β_{1} = \dots = β_{m} = 0$ ” in formula (4). Then, the Lagrange multiplier ( $L M$ ) statistic and F-statistic ( $L M_{F}$ ) are constructed in the auxiliary regression to perform the tests:

L M = T N (S S R_{0} - S S R_{1}) / S S R_{0}

(5)

\begin{array}{l} L M_{F} = [(S S R_{0} - S S R_{1}) / m K] / \\ [S S R_{0} / (T N - N - m K)] \end{array}

(6)

where $S S R_{0}$ denotes the panel sum of the squared residuals under the null hypothesis (linear panel model with individual effects), $S S R_{1}$ denotes the panel sum of the squared residuals under the alternative hypothesis (PSTR model with two regimes), and $K$ is the number of the explanatory variables. Under the null hypothesis, the $L M$ statistics have an asymptotic $X^{2} (m K)$ distribution, whereas the $L M_{F}$ statistics are asymptotically distributed as $F (m k, T N - N - m K)$ .

If the null hypothesis of linearity, i.e., the null hypothesis of no transition function ( $H_{0} : r = 0$ , where $r$ is the number of transition functions in the PSTR model), is rejected, then we must determine the appropriate number of transition functions to adequately capture all of the nonlinearity (Heidari, Katircioğlu, and Saeidpour 2015). A remaining nonlinearity test proposed by Fouquau, Hurlin, and Rabaud (2008) can help to test whether there is one transition function ( $H_{0} : r = 1$ ) or at least two transition functions ( $H_{1} : r = 2$ ). Thus, we may consider a PSTR model with two transition functions:

\begin{array}{l} Y_{i t} = μ_{i} + θ X_{i t} + β_{0} P S_{i t} + β_{1} P S_{i t} g_{1} (q_{i t}; γ_{1}, c_{1}) \\ + α_{1} P S_{i t} g_{2} (q_{i t}; γ_{2}, c_{2}) + ε_{i t} \end{array}

(7)

Similarly, the second transition function $g_{2} (q_{i t}; γ_{2}, c_{2})$ is replaced by its first-order Taylor expansion around $γ_{2} = 0$ . The remaining nonlinearity test is equivalent to testing the null hypothesis $H_{0} : α_{1} = 1$ . If it is not rejected, then the procedure ends; otherwise, we have to test whether there is a third transition function, and so on.

Another key issue is the selection of an appropriate value for $m$ . A sequence of tests for choosing between $m = 1$ and $m = 2$ proposed by A. González, Teräsvirta, and van Dijk (2005) is as follows. In the auxiliary regression (4), we can test the null hypotheses $H_{03} : β_{3} = 0$ , $H_{02} : β_{2} = 0 | β_{3} = 0$ , and, in turn, $H_{01} : β_{1} = 0 | β_{3} = β_{2} = 0$ . If the rejection of $H_{02}$ is the strongest, we should select m=2; otherwise, we should select m=1.

In Table 3, we present the results of the linearity tests and remaining nonlinearity tests. The null hypothesis of linearity ( $H_{0} : r = 0$ ) is rejected at the 1% significance level for all models. The results reveal the existence of a nonlinear relationship between the demographic structure and outbound tourism demand. Thus, the null hypothesis of no remaining nonlinearity ( $H_{0} : r = 1$ ) cannot be rejected, which means that there is only one transition function.

Table 3.

Tests for Linearity and Remaining Nonlinearity in the PSTR Model.

Model	Dependent Variable	Explanatory Variable	Transition Variable	Statistic	H₀: r = 0 vs. H₁: r = 1	H₀: r = 1 vs. H₁: r = 2	Decision
1	LnOTR	LnPP	LnGNI	LM	44.705***	1.245	r = 1
1	LnOTR	LnPP	LnGNI	LM_F	14.902***	0.082	r = 1
2	LnOTR	LnEDR	LnGNI	LM	21.588***	1.184	r = 1
2	LnOTR	LnEDR	LnGNI	LM_F	7.196***	0.061	r = 1
3	LnOTR	LnPFP	LnGNI	LM	33.929***	0.905	r = 1
3	LnOTR	LnPFP	LnGNI	LM_F	11.310***	0.302	r = 1
4	LnOTR	LnPUP	LnGNI	LM	44.826***	1.594	r = 1
4	LnOTR	LnPUP	LnGNI	LM_F	14.942***	0.198	r = 1
5	LnOTR	LnHER	LnGNI	LM	47.114***	1.132	r = 1
5	LnOTR	LnHER	LnGNI	LM_F	15.705***	0.120	r = 1
6	LnOTER	LnPP	LnGNI	LM	90.133***	0.839	r = 1
6	LnOTER	LnPP	LnGNI	LM_F	30.044***	0.061	r = 1
7	LnOTER	LnEDR	LnGNI	LM	36.573***	1.053	r = 1
7	LnOTER	LnEDR	LnGNI	LM_F	12.191***	0.018	r = 1
8	LnOTER	LnPFP	LnGNI	LM	73.426***	1.343	r = 1
8	LnOTER	LnPFP	LnGNI	LM_F	24.475***	0.114	r = 1
9	LnOTER	LnPUP	LnGNI	LM	88.036***	1.202	r = 1
9	LnOTER	LnPUP	LnGNI	LM_F	29.345***	0.067	r = 1
10	LnOTER	LnHER	LnGNI	LM	46.007***	1.051	r = 1
10	LnOTER	LnHER	LnGNI	LM_F	15.336***	0.017	r = 1

***

Significance at 1%.

The results of the sequence of tests for selecting $m$ are reported in Table 4. The null hypotheses of $H_{02} : β_{2} = 0 | β_{3} = 0$ and $H_{01} : β_{1} = 0 | β_{3} = β_{2} = 0$ cannot be statistically rejected, whereas $H_{03} : β_{3} = 0$ is rejected at the significance level of 1%. Consequently, all models with one transition function ( $r = 1$ ) and one location parameter ( $m = 1$ ) are suitable and are selected.

Table 4.

Tests for Selecting m (the Number of Location Parameters).

Model	Dependent Variable	Explanatory Variable	Transition Variable	Statistic	H₀₃^*	H₀₂^*	H₀₁^*	Decision
1	LnOTR	LnPP	LnGNI	LM	312.190***	1.468	0.288	m=1
1	LnOTR	LnPP	LnGNI	LM_F	104.063***	1.156	0.960	m=1
2	LnOTR	LnEDR	LnGNI	LM	369.142***	0.752	0.214	m=1
2	LnOTR	LnEDR	LnGNI	LM_F	123.047***	0.277	0.071	m=1
3	LnOTR	LnPFP	LnGNI	LM	423.467***	0.813	0.256	m=1
3	LnOTR	LnPFP	LnGNI	LM_F	141.156***	0.115	0.085	m=1
4	LnOTR	LnPUP	LnGNI	LM	402.049***	1.752	0.291	m=1
4	LnOTR	LnPUP	LnGNI	LM_F	134.016***	0.733	0.097	m=1
5	LnOTR	LnHER	LnGNI	LM	306.296***	1.375	0.142	m=1
5	LnOTR	LnHER	LnGNI	LM_F	102.099***	0.366	0.047	m=1
6	LnOTER	LnPP	LnGNI	LM	408.629***	1.513	0.225	m=1
6	LnOTER	LnPP	LnGNI	LM_F	136.210***	0.499	0.075	m=1
7	LnOTER	LnEDR	LnGNI	LM	413.618***	2.423	0.077	m=1
7	LnOTER	LnEDR	LnGNI	LM_F	137.873***	0.420	0.026	m=1
8	LnOTER	LnPFP	LnGNI	LM	519.599***	1.088	0.179	m=1
8	LnOTER	LnPFP	LnGNI	LM_F	173.199***	0.877	0.059	m=1
9	LnOTER	LnPUP	LnGNI	LM	506.792***	1.746	0.217	m=1
9	LnOTER	LnPUP	LnGNI	LM_F	168.931***	0.732	0.072	m=1
10	LnOTER	LnHER	LnGNI	LM	417.639***	1.003	0.145	m=1
10	LnOTER	LnHER	LnGNI	LM_F	139.213***	0.994	0.048	m=1

***

Significance at 1%.

Empirical Results

Baseline Results

We employed the simulated annealing method to obtain the initial values of the slope parameter $γ$ and location parameter $c$ . Then, the nonlinear least squares (NLS) method was used to estimate the parameters of the PSTR model. Table 5 presents the parameter estimates of the final PSTR models with one transition function and one location parameter.

Table 5.

Estimated Results for the PSTR Model.

	(LnPP)		(LnEDR)		(LnPFP)		(LnPUP)		(LnHER)
	LnOTR	LnOTER	LnOTR	LnOTER	LnOTR	LnOTER	LnOTR	LnOTER	LnOTR	LnOTER
LnPP	3.588*** (0.542)	3.657*** (0.568)
LnEDR			−1.622*** (0.142)	0.416*** (0.143)
LnPFP					−8.113*** (1.553)	−1.834* (1.466)
LnPUP							0.230* (0.245)	0.410*** (0.239)
LnHER									−0.742*** (0.073)	−0.374*** (0.058)
LnPD	0.341(0.223)	1.159*** (0.220)	1.171*** (0.204)	0.323* (0.204)	0.235* (0.207)	0.411* (0.193)	0.321* (0.220)	0.398* (0.214)	0.492** (0.214)	0.243* (0.208)
RGDP	−0.002(0.003)	−0.009*** (0.003)	−0.004(0.003)	−0.010*** (0.003)	−0.003(0.003)	−0.006*** (0.003)	−0.001(0.003)	−0.010*** (0.003)	−0.002(0.003)	−0.007*** (0.003)
SR	−0.006*** (0.002)	−0.001(0.002)	−0.011*** (0.002)	−0.006*** (0.002)	−0.007*** (0.002)	−0.006*** (0.002)	−0.008*** (0.002)	−0.002(0.002)	−0.006*** (0.002)	−0.002(0.002)
LnEL	1.132*** (0.276)	1.195*** (0.279)	1.634*** (0.283)	0.857*** (0.287)	1.263*** (0.276)	0.863*** (0.257)	1.414*** (0.298)	0.983*** (0.291)	1.544*** (0.288)	0.534** (0.267)
INF	0.004(0.001)	−0.001*** (0.001)	0.007(0.001)	−0.001*** (0.001)	0.004(0.001)	−0.001*** (0.001)	0.004(0.001)	−0.001*** (0.001)	0.006* (0.001)	−0.002* (0.001)
LnOPEN	0.197*** (0.055)	0.303*** (0.053)	0.043(0.053)	0.268*** (0.053)	0.126** (0.055)	0.108** (0.051)	0.153*** (0.057)	0.273*** (0.055)	0.090(0.055)	0.066(0.053)
LnEXCR	−0.068*** (0.019)	−0.080*** (0.019)	−0.056*** (0.019)	−0.063*** (0.019)	−0.026* (0.019)	−0.071*** (0.018)	−0.044** (0.020)	−0.048** (0.019)	−0.058*** (0.019)	−0.079*** (0.019)
Time fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
$γ$	2.609	2.181	1.206	1.645	2.020	0.927	3.276	2.528	1.651	1.031
$c$	5.940	6.625	5.950	5.829	5.944	6.594	6.440	6.624	5.944	6.620
$P S \times g (\cdot)$	0.608*** (0.038)	0.420*** (0.029)	2.082*** (0.094)	1.168*** (0.093)	0.882*** (0.040)	1.112*** (0.043)	0.537*** (0.028)	0.530*** (0.027)	1.234*** (0.061)	1.052*** (0.049)
$e_{i t}$	[3.717, 4.196]	[3.687, 4.077]	[−0.892, 0.456]	[0.480, 1.580]	[−7.879, −7.231]	[−1.550, −0.739]	[0.249, 0.767]	[0.415, 0.932]	[−0.368, 0.492]	[−0.134, 0.668]

Note: (LnPP) represents the model constructed with LnPP as the core explanatory variable, and other models are deduced likewise. LnOTR and LnOTER represent the model constructed with LnOTR or LnOTER as independent variables. *, **, and *** denote significance at 10%, 5%, and 1%, respectively. The standard errors are in parentheses.

Overall, the results indicate that the impact of demographic structure factors on outbound tourism demand shows complicated nonlinear characteristics. For all models, the slope parameters are relatively small, taking values between 0.927 and 3.276. This result implies that the transition function is continuous and smoothly switched between regimes. More precisely, when the transition variable (GNI per capita) increases, the coefficient of demographic structure factors on outbound tourism demand varies steadily. The results provide evidence for how demographic structure factors affect outbound tourism demand. The details are as follows.

(1) Evidence on the age channel. There is a significantly positive relationship between the proportion of the population aged 15–64 years (PP) and outbound tourism demand. Specifically, after GNI per capita exceeds approximately 380 USD (i.e., location parameter c is 5.940) and 754 USD (i.e., location parameter c is 6.625), the coefficient of PP on the outbound tourism rate (OTR) increases from 3.717 to 4.196 and the coefficient of PP on the outbound tourism expenditure rate (OTER) increases from 3.687 to 4.077. That is, when the per capita income is higher, the increase in the proportion of the population aged 15–64 years will lead to greater outbound tourism rates and travel expenditures.

Interestingly, there is a U-shaped relationship between the elderly dependency ratio (EDR) and OTR. When GNI per capita exceeds approximately 384 USD (i.e., location parameter c is 5.950), the coefficient of EDR on OTR changes from −0.892 to 0.456. This change implies that income limits elderly individuals’ participation in travel abroad. If the income constraints are relaxed, then the potential of the elderly travel market will be released, which provides suggestion for dealing with an aging population. In terms of consumption, the results show that EDR promotes the outbound tourism expenditure rate (OTER). This effect is greater (from 0.480 to 1.580) when GNI per capita exceeds approximately 340 USD (i.e., location parameter c is 5.829). Our findings support the view of Jang et al. (2004) that older travelers tend to spend more on outbound tourism.

(2) Evidence on the gender channel. The proportion of the female population (PFP) has a significant negative effect on outbound tourism demand, but this negative effect decreases from −7.879 to −7.231 and from −1.550 to −0.739 as GNI per capita exceeds approximately 381 USD (relative to OTR) and 731 USD (relative to OTER), respectively. The results show that an increase in the female population will restrain outbound tourism demand; however, such restraints will gradually be weakened with an increase in GNI per capita. It is suggested that the female travel market has great potential. Changes in the concept of marriage and the transformation of women’s social roles have led to a significant increase in the number of women without family restraints (Yeoman et al. 2011). They have more leisure time and are inclined to travel. It is foreseeable that women will have more travel opportunities as income levels increase.

(3) Evidence on the urbanization channel. With respect to the coefficients estimated, the income threshold value is 6.440 (i.e., 626 USD, relative to OTR) and 6.624 (i.e., 753 USD, relative to OTER). Below the threshold, an increase in the proportion of the urban population (PUP) of 1% results in increases of 0.249% in OTR and 0.415% in OTER. Above the threshold, the growth of OTR and OTER will eventually be increased to 0.767% and 0.932% through the adjustment of the transition function. In other words, the higher the per capita income, the more the rise in urbanization increases outbound tourism demand. On the one hand, consumption concepts and patterns spread from urban to rural areas through urbanization, which has an impact on the tourism demand of rural residents. On the other hand, urbanization makes urban residents increasingly yearn for different life experiences, such as rural life experience, thus creating more demand for tourism.

(4) Evidence on the education channel. There is also a U-shaped relationship between the higher education ratio (HER) and outbound tourism demand. The income threshold value is 5.944 (i.e., 381 USD, relative to OTR) and 6.620 (i.e., 750 USD, relative to OTER). Below the threshold, an increase in HER of 1% results in decreases of 0.368% in OTR and 0.134% in OTER. Above the threshold, the negative effect of HER is converted into a positive effect through the adjustment of the transition function. For every 1% increase in HER, OTR is increased by 0.492%, and OTER is increased by 0.668%. That is, income effectively regulates the relationship between educational attainment and outbound tourism demand. It is believed that people with a higher level of education are more likely to have a high or stable income than those without higher education; thus, they are more likely to travel abroad (Dai and Sun 2014). Our findings add to this idea that only when higher education is matched with higher income can outbound travel participation and spending be improved.

Among all the control variables, we find that population density (PD), the employment level (EL), and the degree of openness (OPEN) have a significant role in promoting outbound tourism demand. In contrast, the savings rate (SR) and the exchange rate of currency (EXCR) have a negative impact on outbound tourism demand. Notably, the impact of the inflation rate (INF) on OTR and OTER varies, indicating that the price level of a country or region impacts only the final expenditure of outbound tourism, not the decisions to travel. RGDP has a slight negative impact on outbound tourism demand; to some extent, this result implies that the economic growth rate of country or region cannot be directly converted into a driving force of the development of the outbound tourism market.

Robustness Tests

To check the robustness of the empirical results, we conducted two additional sets of estimations. First, we introduce the panel threshold regression (PTR) model to check the identified nonlinear features. GNI per capita is used as a threshold variable, and the regression model is constructed as follows:

\begin{array}{l} Y_{i t} = θ X_{i t} + α_{1} P S_{i t} I (G N I_{i t} \leq q_{i t}) + \\ α_{2} P S_{i t} I (G N I_{i t} > q_{i t}) + μ_{i} + ε_{i t} \end{array}

(8)

where $q_{i t}$ denotes the threshold variable and $I (\cdot)$ denotes the transition function with an indicator function. $α_{1}$ and $α_{2}$ represent the coefficients of demographic structure variables ( $P S_{i t}$ ) on outbound tourism demand ( $Y_{i t}$ ) when GNI_it ≤ q_it and $G N I_{i t} > q_{i t}$ , respectively. $X_{i t}$ is a set of control variables that shape outbound tourism demand. μ _i represents the fixed individual effect, and $ε_{i t}$ is the error term.

The results are reported in Table 6. For each model, we consider cases both including and excluding the control variables, which not only reflects the robustness of the estimation results but also isolates the contribution of the demographic structure variables to the model fit. The results confirm that demographic structure factors have a significant impact on outbound tourism demand. More precisely, there is a nonlinear relationship between demographic structure factors and outbound tourism demand because of the moderating effect of income.

Table 6.

Robustness Test: Estimated Results for the PTR Model.

	(LnPP)		(LnEDR)		(LnPFP)		(LnPUP)		(LnHER)
	LnOTR	LnOTER	LnOTR	LnOTER	LnOTR	LnOTER	LnOTR	LnOTER	LnOTR	LnOTER
Without control variables
GP_1	6.356*** (0.466)	12.921*** (0.473)	−0.209* (0.127)	1.130*** (0.152)	−8.957*** (1.888)	−3.778*** (2.319)	2.088*** (0.227)	3.006*** (0.287)	−0.015* (0.062)	−0.652*** (0.051)
GP_2	6.665*** (0.458)	13.039*** (0.471)	0.994*** (0.120)	1.604*** (0.150)	−8.502*** (1.886)	−3.573*** (2.320)	2.582*** (0.221)	3.182*** (0.283)	0.550*** (0.040)	0.907*** (0.045)
R²	0.507	0.561	0.462	0.485	0.401	0.412	0.478	0.458	0.498	0.593
With control variables included
GP_1	3.914*** (0.533)	6.179*** (0.537)	−0.493*** (0.163)	1.290*** (0.168)	−7.678*** (1.683)	−1.425* (1.770)	0.308*(0.242)	0.203* (0.258)	−0.219*** (0.063)	−0.187*** (0.069)
GP_2	4.210*** (0.527)	6.330*** (0.532)	0.722*** (0.158)	1.486*** (0.162)	−7.321*** (1.682)	−1.263* (1.772)	0.753*** (0.239)	0.348* (0.258)	0.258*** (0.052)	0.238*** (0.052)
LnPD	1.162(0.226)	1.353*** (0.234)	0.620*** (0.217)	0.407* (0.224)	0.512** (0.215)	0.476** (0.233)	0.438** (0.218)	0.386(0.234)	0.437* (0.227)	0.518** (0.228)
RGDP	−0.001(0.003)	−0.010*** (0.003)	0.000(0.003)	−0.009*** (0.003)	−0.001(0.003)	−0.008** (0.003)	−0.001(0.003)	−0.011*** (0.003)	−0.001(0.003)	−0.011*** (0.003)
SR	−0.006*** (0.002)	−0.003(0.002)	−0.003(0.002)	−0.007*** (0.002)	−0.005** (0.002)	0.000(0.002)	−0.004* (0.002)	−0.006*** (0.002)	−0.002(0.002)	−0.008*** (0.002)
LnEL	0.898*** (0.280)	0.502* (0.293)	0.584** (0.291)	0.440(0.300)	0.577** (0.285)	0.201(0.307)	0.750*** (0.288)	0.210(0.310)	0.821*** (0.298)	0.177(0.299)
INF	0.004(0.001)	−0.002*** (0.001)	0.004(0.001)	−0.002*** (0.001)	0.004(0.001)	−0.002*** (0.001)	0.004* (0.001)	−0.003** (0.001)	0.005* (0.001)	−0.002** (0.001)
LnOPEN	0.164*** (0.055)	0.315*** (0.057)	0.115** (0.056)	0.264*** (0.059)	0.096* (0.057)	0.239*** (0.061)	0.135** (0.056)	0.302*** (0.061)	0.132** (0.058)	0.303*** (0.059)
LnEXCR	−0.064*** (0.019)	−0.081*** (0.020)	−0.040** (0.020)	−0.059*** (0.021)	−0.015(0.020)	−0.034(0.022)	−0.038* (0.020)	−0.037* (0.021)	−0.038* (0.020)	−0.047** (0.020)
Time fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
R ²	0.619	0.735	0.598	0.721	0.600	0.699	0.600	0.701	0.571	0.720

Note: GP_1 and GP_2 are the per capita income ranges classified by the threshold value. *, **, and *** denote significance at 10%, 5%, and 1%, respectively. The standard errors are in parentheses.

Second, according to the results and analyses of the PSTR model, the relationships between demographic structure factors and outbound tourism demand are different for countries and regions with high and low income levels. To confirm this difference, we divide the sample countries and regions into two groups of different income levels based on the calculated location parameters ( $c$ ). Then, we employ a fixed effect model with robust standard errors, the Driscoll-Kraay standard errors (DKSE)⁴ provided by the nonparametric covariance matrix estimation method, to reinvestigate the nexus between demographic structure factors and outbound tourism demand. For the reasons mentioned above, we still consider cases both including and excluding the control variables.

The results are reported in Table 7 and show that the estimated coefficients of the demographic structure variables are roughly consistent with the results of the PSTR model. For different income groups, the estimated coefficient of the demographic structure variables in the low-income group is smaller than that in the high-income group. Thus, our results once again confirm the existence of a nonlinear nexus and indicate that the income level moderates the effect of the demographic structure on outbound tourism demand.

Table 7.

Robustness Test: Estimated Results Grouped by Location Parameters.

	LnPP				LnEDR				LnPFP				LnPUP				LnHER
LnOTR	G1	G1	G2	G2	G1	G1	G2	G2	G1	G1	G2	G2	G1	G1	G2	G2	G1	G1	G2	G2
LnPP	6.729** (0.494)	3.703*** (0.226)	8.357*** (2.092)	9.430*** (4.181)
LnEDR					−1.571** (0.180)	−2.001*** (0.264)	0.267** (0.881)	2.896*** (0.480)
LnPFP									−5.109* (3.058)	−3.714** (3.035)	−3.316*** (7.772)	−2.749** (10.035)
LnPUP													3.451** (0.148)	1.494*** (0.143)	4.459*** (0.478)	7.601*** (0.882)
LnHER																	−0.698** (0.038)	−0.291*** (0.059)	1.104*** (0.199)	0.497* (0.392)
Control variables	No	Yes	No	Yes	No	Yes	No	Yes	No	Yes	No	Yes	No	Yes	No	Yes	No	Yes	No	Yes
R ²	0.199	0.439	0.664	0.825	0.191	0.403	0.207	0.7935	0.262	0.4034	0.272	0.7735	0.2073	0.4273	0.1501	0.8458	0.2803	0.4214	0.4175	0.7684
	LnPP				LnEDR				LnPFP				LnPUP				LnHER
LnOTER	G1		G2		G1		G2		G1		G2		G1		G2		G1		G2
LnPP	12.372** (1.058)	4.766*** (0.379)	14.989*** (1.402)	6.423*** (0.596)
LnEDR					2.741*** (0.526)	1.167*** (0.310)	4.436*** (1.214)	1.307*** (0.398)
LnPFP									−3.761** (6.340)	−4.516** (3.557)	−1.935* (5.789)	−1.155* (4.693)
LnPUP													5.088** (0.164)	0.190* (0.231)	5.574*** (0.415)	2.788*** (0.313)
LnHER																	−1.331*** (0.084)	−0.425*** (0.047)	0.961*** (0.095)	0.502*** (0.095)
Control variables	No	Yes	No	Yes	No	Yes	No	Yes	No	Yes	No	Yes	No	Yes	No	Yes	No	Yes	No	Yes
R ²	0.326	0.687	0.684	0.947	0.232	0.653	0.200	0.944	0.200	0.638	0.284	0.943	0.208	0.638	0.392	0.952	0.444	0.655	0.522	0.949

Note: G1 and G2 are the per-capita income ranges classified by the location parameter $γ$ . *, **, and *** denote significance at 10%, 5%, and 1%, respectively. The standard errors are in parentheses.

Discussion

In recent decades, we have seen a rapid increase in international tourism demand. According to UNWTO (2019) statistics, international tourist arrivals worldwide increased 6% in 2018 to 1.4 billion. Meanwhile, international tourism is also playing an increasing role in national and global economies. Identifying and analyzing the determinants of tourism demand are critical to understanding and interpreting the changes in the past and predicting possible paths for future tourism development (Peng et al. 2015). Given this importance, a plethora of studies on international tourism demand have been conducted (e.g., Crouch 1992; Lim 1997; Song and Li 2008). Economic factors have been suggested to be the leading determinants of international tourism demand, and income frequently provides the greatest explanatory power (Crouch 1994). However, noneconomic factors are also important. Increasing urbanization, population, migration, education, and personal freedom have all stimulated individuals’ desire for outbound travel. Satisfactory conclusions on the relative importance of economic and noneconomic factors have not yet been reached.

Our research contributes to resolving the debate over the importance of economic and noneconomic factors. We first confirm the crucial impact of income on outbound tourism demand and further explore the moderating effect of income on the relationship between noneconomic factors and outbound tourism demand. Similarly, L. Wang, Fang, and Law (2018) examine the impacts of air quality on outbound tourism demand while considering a moderator of disposable income. Differently, our study demonstrates the existence of the moderating effect of income by using a PSTR model instead of simply grouping by income or introducing cross-terms with income variables. The PSTR model helps to precisely identify the threshold characteristics of income. In addition, the PSTR model has the great advantage of capturing data heterogeneity to improve the accuracy of parameter estimation.

Furthermore, our research provides evidence for the impact of noneconomic factors on outbound tourism demand from a demographic structure perspective.

Strangely, demography has not received sufficient discussion, even though it is clearly relevant to outbound tourism demand. One possible reason is that in economic models, demographic factors can hardly provide satisfactory explanatory power. Therefore, we consider the moderating effect of income while exploring the possible relationship between demographic factors and outbound tourism demand to offer a more reasonable explanation of the role of demography. The results reveal that the natural structure of demographics, such as age and gender, and the social structure of demographics, such as education and urbanization, can shape outbound tourism demand. Additionally, the influence of the demographic structure on outbound tourism demand differs based on the income regime (see Figure 1). Specifically, the cohort of individuals aged 15–64 years dominates the outbound tourism market. As incomes rise, each additional unit in the proportion of the population aged 15–64 years creates more demand for travel abroad. The female and elderly markets have enormous potential, mainly after the relaxation of income restrictions. Increasing urbanization has stimulated individuals’ desire for outbound travel, especially in the case of the higher-income cohort. Additionally, only when higher education is matched with higher income can outbound travel participation and spending be improved. Such findings supplement the existing literature on how demographic effects shape outbound tourism demand in the context of considering income as a moderator.

Figure 1.

The influences of demographic structure factors on outbound tourism demand under different income threshold.

Conclusion

This article aims to explore the relationship between demographic factors and outbound tourism demand while considering the moderating effects of income. The data set covers 72 countries and regions during the 2000–2014 period. Based on a PSTR model, the results suggest that there is a nonlinear relationship between demographic structure factors and outbound tourism demand and that the demographic effects can vary under different income cohorts.

Regarding the demand for outbound tourism, this study highlights that participation and expenditure in outbound tourism are related to age, gender, education, and urbanization.

The cohort of individuals aged 15–64 years old dominates the outbound tourism market. As incomes rise, each additional unit in the proportion of the population aged 15–64 creates more demand for travel abroad.

Income limits the participation of older cohorts in traveling abroad, but they are inclined to spend more during the trip.

An increase in the female population will restrain outbound tourism demand, but such restraints will gradually be weakened with an increase in GNI per capita.

Income effectively moderates the relationship between education and outbound tourism demand. Only when higher education is matched with higher income can outbound travel participation and spending be promoted.

Increasing urbanization has stimulated individuals’ desire to travel abroad, especially in the case of the higher-income cohort.

These conclusions may provide some policy implications for tourism marketers and governments. Tourism marketers should carefully analyze the tourism preferences and behaviors of target markets from the perspective of demographic characteristics and customize the tourism marketing strategy according to the age distribution, gender attributes, location of residence (urban or rural), and educational level of potential consumers. For example, for city dwellers in developed regions, the attraction of a modern city tour may be far lower than that of a trip to a countryside where it is close to nature; for elderly or female cohorts in some developed countries, targeted tourism advertising and the customization of tourism products and services are necessary. Undoubtedly, it can be foreseen that only emphasizing the price advantage of tourism products without further target market segmentation will no longer be competitive. Governments should take measures in advance to cope with the demographic transition that has occurred in most countries. Considering the impact of population aging on future tourism demand, it is necessary for governments to further improve tourism facilities. Elderly people should be provided with targeted facilities, products and services that are different from those of mass tourism to meet or enhance their convenience in tourism activities. For instance, transportation hubs should be constructed with completely barrier-free access (e.g., accessible train stations, barrier-free bus stations, barrier-free airports); the number of barrier-free hotels should be increased, or the facilities of existing hotels should be improved to make them accessible to wheelchair users; barrier-free tours in tourist attractions should be implemented; and service staff with professional nursing abilities should be provided to accompany elderly individuals during the trip.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported by a grant from the National Social Science Fund of China (17XGL012).

ORCID iDs

Songshan (Sam) Huang

Jianping Zha

Notes

Author Biographies

Zhiyong Liis a Professor in the School of Tourism, Sichuan University, China. His research interests focus on tourism economics and ecological economics.

Haoyu Shu, is a PhD student in the School of Tourism, Sichuan University, China. Her research interests focus on tourism economics and ecological economics.

Ting Tan, is a master’s degree candidate in the School of Tourism, Sichuan University, China. His research interests focus on tourism economics.

Songshan (Sam) Huang is a Professor of Tourism and Services Marketing in the School of Business and Law at Edith Cowan University, Perth, Australia. His research interests include tourist behavior, destination marketing, tour guiding, and various Chinese tourism and hospitality issues.

Jianping Zha is an Associate Professor in the School of Tourism, Sichuan University, China. His research interests focus on tourism economics and ecological economics.

References

Andreu

Kozak

Avci

Cifter

2006. “Market Segmentation by Motivations to Travel: British Tourists Visiting Turkey.” Journal of Travel & Tourism Marketing 19 (1): 1–14.

Bernini

Cracolici

M. F.

2015. “Demographic Change, Tourism Expenditure and Life Cycle Behaviour.” Tourism Management 47:191–205.

Bloom

D. E.

Canning

Fink

2010. “The Greying of the Global Population and Its Macroeconomic Consequences.” Twenty-First Century Society 5 (3): 233–42.

Bond

M. E.

Ladman

J. R.

1972. “International Tourism and Economic Development: A Special Case for Latin America.” Review of Financial Economics 8 (1): 43.

Christensen

Doblhammer

Rau

Vaupel

J. W.

2009. “Ageing Populations: The Challenges Ahead.” Lancet 374 (9696): 1196–208.

Collins

Tisdell

2002. “Gender and Differences in Travel Life Cycles.” Journal of Travel Research 41 (2): 133–43.

Crouch

G. I.

1992. “Effect of Income and Price on International Tourism.” Annals of Tourism Research 19 (4): 643–64.

Crouch

G. I.

1994. “The Study of International Tourism Demand: A Review of Findings.” Journal of Travel Research 33 (1): 12–23.

Crouch

G. I.

1996. “Demand Elasticities in International Marketing: A Meta-analytical Application to Tourism.” Journal of Business Research 36 (2): 117–36.

10.

Dai

X. F.

Sun

P. P.

2014. “Non-linear Relationships between Income and Outbound Tourism Rate: Empirical Evidence Based on Threshold Panel Models.” Tourism Tribune 29 (9): 13–23.

11.

De Vita

. 2014. “The Long-Run Impact of Exchange Rate Regimes on International Tourism Flows.” Tourism Management 45:226–33.

12.

Driscoll

J. C.

Kraay

A. C.

1998. “Consistent Covariance Matrix Estimation with Spatially Dependent Panel Data.” Review of Economics and Statistics 80 (4): 549–60.

13.

Duarte

Pinilla

Serrano

2013. “Is There an Environmental Kuznets Curve for Water Use? A Panel Smooth Transition Regression Approach.” Economic Modelling 31:518–57.

14.

Eugenio-Martin

J. L.

Campos-Soria

J. A.

2011. “Income and the Substitution Pattern between Domestic and International Tourism Demand.” Applied Economics 43 (20): 2519–31.

15.

Feyrer

2007. “Demographics and Productivity.” Review of Economics and Statistics 89 (1): 100–9.

16.

Figueroa-Domecq

Pritchard

Segovia-Pérez

Morgan

Villacé-Molinero

2015. “Tourism Gender Research: A Critical Accounting.” Annals of Tourism Research 52:87–103.

17.

Firestone

Shelton

B. A.

1994. “A Comparison of Women’s and Men’s Leisure Time: Subtle Effects of the Double Day.” Leisure Sciences 16 (1): 45–60.

18.

Fleischer

Pizam

2002. “Tourism Constraints among Israeli Seniors.” Annals of Tourism Research 29 (1): 106–23.

19.

Fouquau

Hurlin

Rabaud

2008. “The Feldstein–Horioka Puzzle: A Panel Smooth Transition Regression Approach.” Economic Modelling 25 (2): 284–99.

20.

Frew

E. A.

Shaw

R. N.

1999. “The Relationship between Personality, Gender, and Tourism Behavior.” Tourism Management 20 (2): 193–202.

21.

Gholipour

H. F.

Tajaddini

Al-Mulali

2014. “Does Personal Freedom Influence Outbound Tourism?” Tourism Management 41:19–25.

22.

Glover

Prideaux

2009. “Implications of Population Ageing for the Development of Tourism Products and Destinations.” Journal of Vacation Marketing 15 (1): 25–37.

23.

González

Teräsvirta

van Dijk

2005. “Panel Smooth Transition Regression Models.” Working Paper Series in Economic and Finance No 604, Stockholm School of Economics, Stockholm.

24.

González

Moral

1995. “An Analysis of the International Tourism Demand in Spain.” International Journal of Forecasting 11 (2): 233–51.

25.

Grimm

Heinsohn

Richter

Metzler

2009. The Impact of Demographic Change on Tourism and Conclusions for Tourism Policy. Germany: Federal Ministry of Economics and Technology.

26.

Hansen

B. E.

1999. “Threshold Effects in Non-dynamic Panels: Estimation, Testing, and Inference.” Journal of Econometrics 93 (2): 345–68.

27.

Heidari

Katircioğlu

S. T.

Saeidpour

2015. “Economic Growth, CO₂ Emissions, and Energy Consumption in the Five ASEAN Countries.” International Journal of Electrical Power & Energy Systems 64:785–91.

28.

Hong

G. S.

Morrison

A. M.

Cai

L. A.

1996. “Household Expenditure Patterns for Tourism Products and Services.” Journal of Travel & Tourism Marketing 4 (4): 15–40.

29.

Jang

S. S.

Bai

Hong

G. S.

O’Leary

J. T.

2004. “Understanding Travel Expenditure Patterns: A Study of Japanese Pleasure Travelers to the United States by Income Level.” Tourism Management 25 (3): 331–41.

30.

Jiang

Liu

Song

2017. “Influencing Factors of Outbound Tourism and Their Different Functions in Developing and Developed Economies.” Tourism Tribune 32 (1): 12–21.

31.

Kattiyapornpong

Miller

K. E.

2011. “Social Structure and Psychographic Explanations of Destination Preference of Australians’ Travel to South-East Asia.” Asia Pacific Journal of Tourism Research 16 (1): 39–56.

32.

Kinnaird

Hall

, eds. 1994. Tourism: A Gender Analysis. Chichester, UK: Wiley.

33.

Kinnaird

Hall

1996. “Understanding Tourism Processes: A Gender-Aware Framework.” Tourism Management 17 (2): 95–102.

34.

Kulendran

Wilson

2000. “Is There a Relationship between International Trade and International Travel?” Applied Economics 32 (8): 1001–9.

35.

Kulendran

Witt

S. F.

2003. “Forecasting the Demand for International Business Tourism.” Journal of Travel Research 41 (3): 265–71.

36.

Zhou

2019. “What Are the Impacts of Demographic Structure on CO₂ Emissions? A Regional Analysis in China via Heterogeneous Panel Estimates.” Science of the Total Environment 650:2021–31.

37.

X. R.

Hudson

2013. “The Application of Generational Theory to Tourism Consumer Behavior: An American Perspective.” Tourism Management 37:147–64.

38.

Liddle

2014. “Impact of Population, Age Structure, and Urbanization on Carbon Emissions/Energy Consumption: Evidence from Macro-Level, Cross-Country Analyses.” Population and Environment 35 (3): 286–304.

39.

Lim

1997. “Review of International Tourism Demand Models.” Annals of Tourism Research 24 (4): 835–49.

40.

Lim

1999. “A Meta-analytic Review of International Tourism Demand.” Journal of Travel Research 37 (3): 273–84.

41.

Lin

V. S.

Liu

Song

2015. “Modeling and Forecasting Chinese Outbound Tourism: An Econometric Approach.” Journal of Travel & Tourism Marketing 32 (1/2): 34–49.

42.

Hung

Wang

Schuett

M. A.

2016. “Do Perceptions of Time Affect Outbound-Travel Motivations and Intention? An Investigation among Chinese Seniors.” Tourism Management 53:1–12.

43.

MacKay

K. J.

1997. “Pictorial Element of Destination in Image Formation.” Annals of Tourism Research 24 (3): 537–565.

44.

Mak

Carlile

Dai

2005. “Impact of Population Aging on Japanese International Travel to 2025.” Journal of Travel Research 44 (2): 151–62.

45.

McCleary

K. W.

Weaver

P. A.

Lan

1994. “Gender-Based Differences in Business Travelers’ Lodging Preferences.” Cornell Hotel and Restaurant Administration Quarterly 35 (2): 51–58.

46.

Meng

Uysal

2008. “Effects of Gender Differences on Perceptions of Destination Attributes, Motivations, and Travel Values: An Examination of a Nature-Based Resort Destination.” Journal of Sustainable Tourism 16 (4): 445–66.

47.

Miles

1999. “Modelling the Impact of Demographic Change upon the Economy.” Economic Journal 109 (452): 1–36.

48.

Mottiar

Quinn

2004. “Couple Dynamics in Household Tourism Decision Making: Women as the Gatekeepers?” Journal of Vacation Marketing 10 (2): 149–60.

49.

Nicolau

J. L.

Más

F. J.

2005. “Heckit Modelling of Tourist Expenditure: Evidence from Spain.” International Journal of Service Industry Management 16 (3): 271–93.

50.

Peng

Song

Crouch

G. I.

Witt

S. F.

2015. “A Meta-analysis of International Tourism Demand Elasticities.” Journal of Travel Research 54 (5): 611–33.

51.

Poterba

J. M.

2001. “Demographic Structure and Asset Returns.” Review of Economics and Statistics 83 (4): 565–84.

52.

Prideaux

2005. “Factors Affecting Bilateral Tourism Flows.” Annals of Tourism Research 32 (3): 780–801.

53.

Pritchard

Morgan

Ateljevic

Harris

, eds. 2007. Tourism and Gender: Embodiment, Sensuality and Experience. Oxford: CABI.

54.

Qiu

Zhang

1995. “Determinants of Tourist Arrivals and Expenditures in Canada.” Journal of Travel Research 34 (2): 43–49.

55.

Riker

2014. “Population Aging and the Economics of International Travel.” Tourism Economics 20 (1): 21–38.

56.

Sakai

Brown

Mak

2000. “Population Aging and Japanese International Travel in the 21st Century.” Journal of Travel Research 38 (3): 212–20.

57.

Sangpikul

2008. “Travel Motivations of Japanese Senior Travellers to Thailand.” International Journal of Tourism Research 10 (1): 81–94.

58.

Shan

Wilson

2001. “Causality between Trade and Tourism: Empirical Evidence from China.” Applied Economics Letters 8 (4): 279–83.

59.

Song

2008. “Tourism Demand Modelling and Forecasting—A Review of Recent Research.” Tourism management 29 (2): 203–20.

60.

Song

Witt

S. F.

Fei

2010. “Tourism Demand Modelling and Forecasting: How Should Demand Be Measured?” Tourism Economics 16 (1): 63–81.

61.

Steiger

2012. “Scenarios for Skiing Tourism in Austria: Integrating Demographics with an Analysis of Climate Change.” Journal of Sustainable Tourism 20 (6): 867–82.

62.

Szromek

A. R.

Januszewska

Romaniuk

2012. “Demographic Phenomena and Demand for Health Tourism Services Correlated in Poland.” American Journal of Tourism Management 1 (1): 10–20.

63.

Tugcu

C. T.

2014. “Tourism and Economic Growth Nexus Revisited: A Panel Causality Analysis for the Case of the Mediterranean Region.” Tourism Management 42:207–12.

64.

UNWTO. 2019. “World Tourism Barometer.” https://www-e--unwto-org-443.web.bisu.edu.cn/toc/wtobarometereng/17/1.html (accessed January 2019).

65.

Uysal

McGehee

N. G.

Loker-Murphy

1996. “The Australian International Pleasure Travel Market: Motivations from a Gendered Perspective.” Journal of Tourism Studies 7 (1): 45.

66.

Wang

Fang

Law

2018. “Effect of Air Quality in the Place of Origin on Outbound Tourism Demand: Disposable Income as a Moderator.” Tourism Management 68:152–61.

67.

Wang

Y. S.

2014. “Effects of Budgetary Constraints on International Tourism Expenditures.” Tourism Management 41:9–18.

68.

Wei

Hao

2010. “Demographic Structure and Economic Growth: Evidence from China.” Journal of Comparative Economics 38 (4): 472–91.

69.

WTO (World Tourism Organization) and ETC (European Travel Commission). 2010. Demographic Change and Tourism. Madrid: WTO and ETC.

70.

Yeoman

Hsu

C. H.

Smith

Watson

2011. Tourism and Demography. Oxford: Goodfellow.

71.

Yeoman

Schanzel

Smith

2013. “A Sclerosis of Demography: How Ageing Populations Lead to the Incremental Decline of New Zealand Tourism.” Journal of Vacation Marketing 19 (2): 91–103.

72.

You

O’Leary

J. T.

2000. “Age and Cohort Effects: An Examination of Older Japanese Travelers.” Journal of Travel & Tourism Marketing 9 (1/2): 21–42.

73.

Zimmer

Brayley

R. E.

Searle

M. S.

1995. “Whether to Go and Where to Go: Identification of Important Influences on Seniors’ Decisions to Travel.” Journal of Travel Research 33 (3): 3–10.