Does it Matter? Examining the Impact of China’s Vacation Policies on Domestic Tourism Demand

Abstract

With the domestic tourism market enjoying rapid expansion in the past decade, a number of significant changes in public policy have affected the institutionalization of public holidays and paid vacations in China. This article examines the impact of public policy in shaping domestic tourism development in China and estimates domestic tourism demand by applying a dynamic model using panel data. Three alternative models are applied to a panel data set made up of the ratio of domestic tourist departures in each of the 29 Chinese originating cities between 2001 and 2010. The empirical results exhibit the significant value of the lagged dependent variable on consumer decision and reveal the causal link between domestic tourism demand and consumer, trip-related and policy attributes. The findings clearly indicate that (a) China’s domestic tourism market is maturing, (b) the vacation policy changes adopted in 2007 have had a significant effect in changing domestic tourism demand, and (c) domestic tourism demand has been substituted by an ever-increasing outbound tourism market. Implications are discussed for policy making and destination management.

Keywords

China vacation policy domestic tourism demand panel data dynamic modeling generalized method of moments (GMM) estimators

This study attempts to examine the impact of China’s vacation policy on domestic tourism demand by applying a dynamic model using panel data. Cooper, Scott, and Kester (2006) headlined their article “The Chinese tourism boom” to describe the rapid growth in the inbound and outbound travel market and the significant expansion of domestic tourism in China during the past decade. More than two billion people participated in domestic tourism in China in 2010 (Figure 1), with the holiday frequency of Chinese citizens reaching an average of 2.46 days per Chinese citizen (China National Tourism Administration [CNTA], 1993-2011). Numerous studies have been completed in the past 10 years to illustrate this economic phenomenon in the Chinese tourism market and to attempt to identify the reasons for this extraordinary growth (Lew, Yu, Ap, & Zhang, 2003; Wen & Tisdell, 2001; Zhang, 2003).

Figure 1

Growth of Domestic Tourism in China

Disposable income and increased leisure time have long been regarded as the two most important forces motivating modern tourism demands (Fayos-Sola & Bueno, 2001; York & Zhang, 2010). Since China’s reform and opening-up policies began in 1978, it has witnessed rapid growth in per capita gross domestic product (GDP) in the past three decades, which reached $4,682 in 2010 (National Bureau of Statistics of China [NBSC], 2002-2011). Moreover, the greater availability of leisure time in the past decade has been directly attributed to the country’s exponential tourism growth during the same period—especially its domestic tourism growth (Gao, 2007; Shao, 2008).

China’s vacation policies remained unchanged during the 50 years since the People’s Republic of China was founded in 1949. From 1999 to 2007, however, two major amendments were made to China’s vacation policies. In 1999, the Central Government amended the Regulation on National Festival and Memorial Day Public Holidays that took effect in May of 2000 in an attempt to drive the growth of China’s economy and tourism market. The revised vacation regulation facilitated the establishment of three week-long holidays. The week-long holiday was instituted by shifting the 2-day weekend from the previous week to the next week plus a 3-day public holiday (York & Zhang, 2010). Until 2007, there were three week-long holidays built around three major festivals and public holidays in China: the Spring Festival (generally in February), Labor Day (May 1), and National Day (October 1). These 3 weeks have officially been labeled as “Golden Week” to indicate the enormous golden opportunities they created for domestic demand and consumption and the economic growth they promoted—including growth in the tourism sector. According to the CNTA (2008), the three Golden Weeks accounted for 40% and 33%, respectively, of annual tourist volume and tourism revenues in the Chinese domestic tourism market in 2007. In addition to providing an impetus to the expansion of the tourism industry, the new vacation policies adopted in 1999 also contributed to China’s economic stability during the Asian financial crisis by providing a major economic stimulus to domestic demand and consumption (Zhang, 2008).

However, 6 years after 1999, fierce debates among government officials, tourism experts, economists, folklorists, sociologists, and other interest groups unfolded. Those discussions were aimed at creating an improved vacation policy system. Some evidence suggests that the Golden Week vacation policy system has led to many social problems that have become obstacles to sustainable tourism development. As Wu, Xue, Morrison, and Leung (2012) summarized, the Golden Weeks failed to promote domestic consumption as much as expected. In addition, the impact of the Golden Weeks on the tourism industry was eventually diluted by the temporal redistribution of domestic tourism demand since most Chinese citizens decided to travel during the longer Golden Weeks, thus reducing domestic travel at other times of the year. Some suggested that an incremental increase in tourism revenues from the Golden Week travel was the result of 3 weeks’ worth of increased tourism consumption—but not a broad, substantial development of China’s tourism industry and economy (Cai, 2009). The public began to recognize that the Golden Week system had a very insignificant impact on the country’s tourism industry and economy—not to mention the aforementioned negative impact it had on sustainable development and social issues. Wu et al. (2012) described the various negative effects of the Golden Week system, which disrupted the regular weekday schedule and impeded the normal operation of the entire economy. Additionally, it produced a serious imbalance between supply and demand in the tourism industry, which resulted in other negative outcomes, including crowding at major attractions, price hikes, and declining service quality during the Golden Weeks. All these adverse effects tarnished the destinations’ image because of low customer satisfaction.

Deemed more negative than positive, the 1999 Golden Week vacation policy system was amended again in 2007. Those revisions cancelled 2 days of the “May Day” vacation and added three new 1-day holidays on traditional Chinese festivals in April, June, and September, beginning in 2008. The new Regulation on National Festival and Memorial Day Public Holidays issued by the Central Government in December 2007 revised the three Golden Weeks to two Golden Weeks plus five shorter holidays. Though the May Day Golden Week was cut short, the total number of annual public holidays increased by 1 day because of the addition of new holidays (distributed over other months). The amended vacation policy in 2007 was thus intended to reduce the temporal concentration of tourism demand for the three Golden Weeks by redistributing them over more months in a year (Table 1). At the same time, the government released the Ordinance on Paid Vacations for Employees, entitling all employees in China to annual paid vacations. This reform constituted the most fundamental change in China’s tourism policy in the past decade and is said to be an important indicator of the nature and direction of vacation policy changes to come (Wu et al., 2012).

Table 1

China’s Public Holiday Policy in 1999 and 2007

1999 Public Holiday Policy	Date/Days	2007 Public Holiday Policy	Date/Days
New Year holiday	January 1/1 day	New Year holiday	January 1/1 day
Spring Festival holiday^a	January or February/3 days	Spring festival holiday^a	January or February/3 days
May Day holiday	May 1/3 days	Qingming festival holiday^a	April/1 day
		May Day holiday	May 1/1 day
		Dragon Boat Festival holiday^a	June/1 day
		Mid-Autumn festival holiday^a	September/1 day
National Day holiday	October 1/3 days	National Day holiday	October 1/3 days
Total	10 days	Total	11 days

These traditional Chinese festivals are determined by the Chinese lunar calendar. So the dates vary yearly. The short 1-day public holiday is now bundled with an adjacent 2-day weekend to make it a 3-day holiday. The Golden Week holiday institutionalized by the 1999 policy has in fact only 3-day legal holidays. The extended 1-week (7-day) public holiday is structured by including the two weekends before and after the three legal holidays. However, people have to work either a 6-day week or 7-day week schedule before or after the Golden Week holiday.

Recognizing the significant changes recently experienced in the vacation policy system in the Chinese domestic tourism market as a whole, this article aims to examine the impact of vacation policy change on domestic tourism demand in China between 2001 and 2010, particularly the effect of the amended vacation policy in 2007 on redistributing demand from four public holidays (three Golden Weeks plus New Year’s holiday) to seven public holidays of varied lengths in different month of the year. It attempts to develop a dynamic model that quantifies the relatively important determinants, income and time, in final demand. Therefore, the main objective of this article is to analyze the economic determinants of domestic tourism demand in China. Its contribution is to provide empirical evidence for public policy decisions in sustainable domestic tourism development.

Literature Review

Tourism Demand and Vacation Policy

According to Burkart and Medlik (1981), the determinants of tourism demand are those factors at work in any society that drive and set limits on the volume of a population’s demand for holiday and travel. Vanhove (2011) indicated tourism demand is determined by a myriad of economic, social, and political factors. The most important set of determinants of tourism demand, however, are economic factors and, more specifically, the amount of disposable income and free time of the population of the generating markets.

Schor (1991) indicated that vacation policies are a natural product of both the economic prosperity brought on by the first industrial revolution and the political enhancement that has come with state recognition of holidays since the late 1700s. Regulations governing public holidays are relatively politically uncontroversial and stable in developed countries (Esping-Anderson, 1990). They also reflect the political systems of the countries they are implemented in, in terms of power arrangements and the imposition of a dominant political ideology (Perry, 2004).

A growing trend toward a more globalized society in recent decades has also resulted in international organizations paying more attention to this issue. For example, in 2003, the United Nations World Tourism Organization (UNWTO) released a number of documents with substantial sections on leisure time (O’Rourke, 2003; Wick, 2001). Those documents focused extensively on the legitimate right of individuals to a reasonable amount of leisure time and emphasized the relationship between leisure time and tourism development. Although the leisure time provisions included in the UNWTO’s documents are not enforceable on signatory nations, they do serve as valuable reference points for countries concerned with formulating leisure time-related regulations (York & Zhang, 2010). Many countries guarantee the right to leisure time as a part of discretionary leisure in their vacation policies.

In modern society, both the public holidays and the paid vacation days are important aspects of the vacation policy system. Most countries provide 8 to 16 days of public holidays and at least 10 to 30 paid vacation days each year (Table 2). Esping-Anderson (1990) described the paid vacation arrangement as a part of the social welfare system in modern Western societies. In many developed economies, employees have 4 to 6 weeks of paid vacation each year outside of public holidays. An increased amount of paid vacation days not only increased the number of individuals who took time off on holidays, but also enabled individuals to arrange their holidays more flexibly, with a shorter “main” holiday and two or more additional “short” holidays each year. Vanhove (2011) suggested that more flexible working time has a direct impact on off-peak holidays and stimulates the staggering arrangement of tourist demand throughout a year. In China, however, the Ordinance on Paid Vacations for Employees was not released until 2007. The paid vacation of 5 days is also markedly less than that of many other countries. The lack of paid vacation days compels most Chinese to concentrate their travel during public holidays, such as the Golden Weeks, which inevitably results in the negative impact of overcrowding and congestions.

Table 2

Amount of Public Holidays and Paid Vacation per Year in Selected Countries (2011)

Country	Public Holidays	Paid Vacation (days)^a
China	11	5
Austria	13	25
Brazil	11	30
Canada	9	10
Denmark	10	25
Finland	10	30
France	10	30
India	16	12
Korea	15	19
New Zealand	11	20
Poland	10	26
Russia	12	28
Singapore	11	14
Switzerland	9	20
United Kingdom	8	28
United States	10	15

Source: Compiled from various publications by United Nations World Tourism Organization (2011).

The number of paid vacation days is the legally mandated minimum by each country.

While most agree that discretionary income and time are two of the most important determinants of tourism demand, many ignore the complicated relationships between income, time, and demand. On the one hand, the money that a person earns in any given year may motivate their travel plans for the following year: this is called “time lag” (Song & Lin, 2010). On the other hand, optimistic consumers who anticipate growth in their income sometimes consume tourism products in advance: this is called “time lead” (Song & Lin, 2010). It is important to note that, at present, the marginal utility of free time is decreasing as many people do not have the funds to take more (short) holidays each year. Free time is, however, still a determining factor in developing countries, and even in many developed countries, such as Japan and the United States (Vanhove, 2011).

An important indicator of tourism demand is gross holiday propensity, also called gross trip participation, which expresses the total number of trips taken as a percentage of the total population during a period of 1 year. The indicator shows the percentage of tourist departures based on the total population. Under a relatively well-institutionalized and stable vacation policy system, the gross holiday propensity of some developed countries exceeds 100%, often approaching 130% and even 180% (Vanhove, 2011).

In China, the gross holiday propensity has exceeded 100% since 2006 and increased to 157.4% in 2010. For some Chinese citizens, that number even approached 246%, which means, on average, each Chinese citizen took 2.46 holidays in 2010 (CNTA, 2011).¹ According to Vanhove’s (2011) definition, the holiday frequency of some Chinese citizens reached 2.46 in 2010. However, there is no doubt that different vacation policies not only influence holiday propensity and frequency, but also affect holiday timing. Though the Central Government of China released the Ordinance on Paid Vacations for Employees, calling on all Chinese workers to take annual paid vacations, most domestic tourists still strongly depend on public holidays—especially the Golden Weeks. Instead of scheduling travel plans throughout the whole year, tourists are more likely to spend their money and time during two or three Golden Weeks. Most individuals are then possibly money-rich but time-poor during the rest of the year.

Most of the current research on the Golden Week vacation policy system’s relation to tourism demand is focused on areas such as microlevel problems that occur during Golden Week vacations and regulatory enhancements made by government agencies during those holidays (Chen & Li, 2007; Ren, 2007; Yang, 2005). Acknowledging the hot debate over further policy changes related to the Golden Week vacation policy system, York and Zhang (2010) examined some of the key dimensions of the policy-making process behind the development of China’s vacation policy system in 1999 and 2007. On top of new social policies, those changes are also attributed to inviting more public participation in the policy-making process, especially in 2007. Wu et al. (2012) were among those who turned their attention to some of the unfolding changes that occurred after the New Vacation Policies were released in 2007, which included cuts to the length of the May Day Golden Week, adding three new, 1-day holidays, and entitling employees to annual paid vacations. In their study, Wu et al. (2012) addressed China’s holiday system reform process by focusing on online news stories between 2005 and 2009. They found that economic stability and consumer rights were the most significant concerns during the 5-year debate (which actually continued 2 years after the release of the New Vacation Policies in 2007). They also found that the political environment in China was opening up to the idea of accepting more public opinion.

It has been several years since the Chinese vacation policy system was reformed. Those reforms represent the most radical change in China’s tourism system in the past decade—not to mention the most obvious forecast of the changing direction of vacation policy. Wu et al. (2012) have pointed out that there has been a dearth of research examining the empirical outcomes of China’s reformed vacation policy system as they relate to the country’s domestic tourism demands. Because China is a developing country, income is still a key factor in determining individuals’ tourism demand. Tourist consumption is mainly dependent not on the structure of the Golden Week, but citizens’ discretionary income (Cai, 2009). On top of the amazing, rapid growth of the Chinese domestic tourism market, the interaction effects between income and time for tourism must also be considered. It is important to examine personal income distribution, the “time lag” between tourism expenditure and income creation, and the “time lead” demand related to income expectations (Song, Witt, & Li, 2009; Vanhove, 2011). This study examines all the above factors and their relationships within a dynamic tourism demand model.

Dynamic Model for Tourism Demand

Static models, which state that the current value of tourism demand is related only to the current values of the explanatory variables, are presented in many early tourism demand studies (Artus, 1972; Gray, 1966; Kwack, 1972; Loeb, 1982; Martin & Witt, 1988; Summary, 1987; Var, Mohammad, & Icoz, 1990; Witt & Martin, 1987). Most of the published studies on causal tourism–demand models before the 1990s were classical static models with very limited diagnostics reported (Garín-Muñoz, 2007). These traditional approaches to tourism demand modeling constructed regression with ordinary least square (OLS) as the main estimation procedures, mainly using linear or power single-equation models and time series data (Song et al., 2009; Song, Dwyer, Li, & Cao, 2012). However, the error terms in static tourism–demand models have generally been found to be highly autocorrelated, and this indicates that the demand relationships are likely to be spurious and that the normal t and F statistics are invalid (Song et al., 2009).

Although static models were still used until 2000 (Akis, 1998; Font, 2000; Lim, 1997; Qu & Lam, 1997), some researchers and practitioners (Little, 1980; Witt, 1980a, 1980b), began to introduce dynamic effects, such as lagged values of dependent and independent variables, into tourism demand models in an attempt to avoid potential problems, such as structural instability, forecasting failures and spurious regression. These models are called autoregressive distribution lag models (ADLM). ADLMs find alternative tourism demand specification by considering the possibility of a change in habit persistence and the influences of social factors, including cultural status, personal preferences and expectations, as referenced by Song et al. (2009). Under different restrictions, the ADLM can produce several types of variation models, such as Static, Autoregressive (AR), Growth Rate, Leading Indicator, Partial Adjustment, Common Factor, Finite Distribute Lag, Dead Start and Error Correction (Croes & Vanegas, 2005; Crouch, Schultz, & Valerio, 1992; Di Matteo & Di Matteo, 1993; Dritsakis & Athanasiadis, 2000; Garín-Muñoz, 2006, 2007, 2009; Ismail et al., 2000; Jensen, 1998; Kliman, 1981; Kuo, Chen, Tseng, Ju, & Huang, 2008; Lim, 2004; Martin & Witt, 1988; Morley 1997, 1998; Song & Witt, 2003; Song & Wong, 2003; Wang, 2009; Witt, 1980a, 1980b).

As several researchers have suggested, tourism has a great deal of inertia, and the dynamic model has the ability to capture the effect of past tourism (Garín-Muñoz & Pérez-Amaral, 2000; Kuo et al., 2008). Garín-Muñoz (2007) indicated that the estimation of the influence of other relevant variables would be overestimated (as the estimated coefficients will involve direct and indirect effects) if the impact of previous tourism were neglected. Therefore, including previous dependent variables in a dynamic model for studying tourism demand is one of the possible strategies for handling the dynamic structure of consumer preferences where changes in taste are regarded as endogenous (Garín-Muñoz, 2007; Kuo et al., 2008).

While making travel decisions, tourists are constrained by money and time and make determinations based on preferences, habits, expectations, and “word of mouth” simultaneously. To provide deep insight into the complex decision-making process of tourists, more studies on the most decisive factors related to the tourism industry must be performed. With that in mind, this study intends to establish a dynamic demand model of domestic tourism with panel data from 2001 to 2010 and including a special focus on analyzing the time effects (presented by different vacation policies) on domestic tourism demand. The income effects and the interactive relationship between these two effects will also be considered in this model. This model will provide relevant information that may serve as a reference for policy making and strategic planning for the domestic tourism industry. Because of the rigidity of supply and the interaction among inbound and outbound tourism markets, this study also contributes valuable information to the international tourism industry.

Model Specification and Econometric Method

In the design of our models, two distinctive features are considered. One is the dynamic specification that takes into account that past demand levels may affect the current number of departures, and the other is the utilization of a panel data set.

Variables of Interest and Data Resources

Data Collection

Based on the available statistical resources in China, 29 originating cities were selected as the sample cities for this study, including 15 cities in economically advanced eastern China (e.g., Shanghai, Hangzhou), 7 central cities (e.g., Wuhan, Changsha), and 7 western interior cities (e.g., Xi’an, Lanzhou). Annual data from 2001 to 2010 were collected from The Yearbook of China Tourism Statistics, an official publication by the CNTA (2002-2011) and The China Statistical Yearbook, an official publication by the NBSC (2002-2011). The time span in the current study includes two one-off events, the SARS event in 2003 and the China vacation policy reform in 2007. The study employs the dynamic panel data of 29 origination cities to examine the effects of vacation policy reform to domestic tourism demand in China. The STATA v.11.2 econometric software is used to obtain the estimation results and analyze the data.

Dependent Variable

The overwhelming majority of current research on tourism demand is focused on international tourism demand. According to the available theory and data, the most frequently used method of presenting tourism demand is via statistics related to tourist arrivals/departures from an originating country to a destination country or tourism receipts/expenditures by visitors from the originating country in the destination country (Barry & O’Hagan, 1972; Croes & Vanegas, 2005; Di Matteo & Di Matteo, 1993; Ismail et al., 2000; Jensen, 1998; Li, Song, & Witt, 2006; Lim, 2004; Morley, 1997, 1998; Song & Witt, 2003; Summary, 1987; Uysal & Crompton, 1985).

The research objective of this study is to measure domestic tourism demand in China. The available data from the government database is the ratio of tourist departures, which is calculated by dividing the number of tourist departures by the resident population. In this particular study, China’s domestic demand for tourism is measured by the ratio of tourist departures in the 29 originating cities included in the research. These 29 cities are the major originating domestic markets in China and CNTA has been tracking departures from these cities since 1994. The ratio of tourist departures takes into account not only the number of tourists but also the population scale. So, it is a better way to measure domestic tourism demand in China than the absolute number of tourist departures or expenditures. The annual ratio of tourist departures was collected from the Yearbook of China Tourism Statistics published by CNTA from 2002 to 2011 for the 29 cities.

Explanatory Variables

Prior literature suggests several possible variables for inclusion in a model to measure international demand, including income, relative prices, exchange rates, and transportation costs. In this study, however, we are going to consider a dynamic specification of domestic tourism. For the domestic tourism market, the most important stimulating factors are economic factors, such as income, time, cost, policy, and one-off events (Vanhove, 2011). Thus, the following explanatory variables are selected for this study.

Income

In order to include income in the demand model, several measures have been used in different empirical studies. Lim (1997) argued that discretionary income, defined as the income remaining after spending on necessities in the region of origin, should be used as the relatively satisfactory measure of income in the demand model. However, these data cannot be easily obtained in practice, especially at the city level (Crouch et al., 2007). Therefore, alternative measures of income have to be used as proxies for discretionary tourist income. One of those alternative measures, the GDP measured in per capita terms, is used in this study. Several prior studies have used per capita GDP as a proxy for the discretionary income level of tourists (Garín-Muñoz, 2007; Ouerfelli, 2008; Rodríguez, Martínez-Roget, & Pawlowska, 2012). The 10-year panel data of per capita GDP in this study were collected from The China Statistical Yearbook between 2002 and 2011 for the 29 originating cities.

Tourism price

Price is an important economic determinant for tourism demand to explain the travel cost (Vanhove, 2011). However, the selection of a tourism price variable included in most studies is particularly difficult (Garín-Muñoz, 2007). Examining the product such as domestic tourism in this study, the available data of tourist demand defined as the ratio of tourist departures are determined from each originating city. The total tourism price, therefore, is difficult to be modeled in the estimation. But based on the Chinese tourist consumption structure, the cost of intercity transportation would normally account for a significant portion of the total price (Gu & Wu, 2003; Li & Xia, 2004; Sun, 2009; Wang et al., 2012). Moreover, Morley (1994) had confirmed that it is reasonable to use the consumer price index (CPI) as a proxy for tourism prices in demand models. In this study, therefore, the CPI of intercity transportation has been established as the proxy variable of tourism price. The annual data on the CPI of intercity transportation were collected from The China Statistical Yearbook between 2002 and 2011 for the 29 cities.

Transport capacity

In the tourism system, transportation is the basic bridging element that connects tourism supply at destinations with tourism demand from originating regions (Vanhove, 2011). With the implementation of the Golden Week vacation policy in China, the travel boom of the Golden Weeks was fueled by citizens from all the originating cities. However, because of great disparity of regional economic development, the gap in transport capacity of the originating cities is widening. Tourists departing from some originating cities have to face serious overcrowding problems that could result in delays, cancellations, and discouraging latent demand, thus curbing the potential outflow of tourists to desired destinations. Therefore, the transport capacity of the originating cities has become a vital factor that affects travelling decisions of urban residents. It is thus necessary to include transport capacity as a proxy variable that measures the largest available number of passengers departing the originating cities by air, rail, land, and water transportation.² Though this measure includes nontourist movement, it can represent the maximum facility capacity for tourists to some extent. The annual data on the number of total passengers in the originating cities were collected from the China Statistical Yearbook between 2002 and 2011 for the 29 cities.

One-off events

Dummy variables are often used to capture the effects of various one-off events, such as the two oil crises in the 1970s (Han, Durbarry, & Sinclair, 2006; Witt, Song, & Louvieris, 2003); the financial crisis in 1997 and the SARS outbreak in 2003 in Asia (Lim, 2004; Wong, Song, Witt, & Wu, 2007); the terrorist attack in the United States in 2001 (Wong et al., 2007); the foot-and-mouth disease outbreak in the United Kingdom in 2001 (Eugenio-Martin et al., 2005); and the various Olympic Games in their host countries (Kim & Song, 1998).

Vacation policy change

A dummy variable, D2007, is used for controlling the effects of the vacation policy changes in China since 2008. It takes the values of 1 for the year 2008, 2009, 2010, and 0 for the rest of the previous years.

D2003

This variable is included to control the effect of the SARS outbreak in 2003. It takes the value of 1 for the year 2003 and 0 for the rest of the time.

Lagged dependent variable

Tourists’ expectations and habit persistence (stable behavior patterns) are usually incorporated in tourism demand models through the use of a lagged dependent variable that is an autoregressive term (Song et al., 2009; Witt, 1980a). There is much less uncertainty associated with holidaying for individuals who prefer travel in developed areas. Furthermore, demonstration effect and “word of mouth” recommendations could, to some extent, play an active role in maintaining the number of departures from originating cities. In general, the path dependence of consumer preference makes the ratios of tourist departures in any year dependent on the numbers in previous years. Song et al. (2009) also indicated further justification for the inclusion of a lagged dependent variable in tourism demand functions by suggesting that there may be a partial adjustment mechanism postulated to allow for rigidities in supply. In other words, once the domestic tourism market has formed, its scale is unlikely to reduce rapidly. Thus, in the dynamic model of this study, the lagged dependent variable must be interpreted as displaying habit persistence and being path dependent. Several studies include the lagged dependent variable for explaining tourism demand (Garín-Muñoz, 2007; Garín-Muñoz & Montero-Martín, 2007; Kuo et al., 2008; Rodríguez et al., 2012; Wang, 2009).

Function Form

The function form describes the relationship between the quantity of tourism demand and its determinants. Though the linear relationship is the simplest one for estimation and interpretation (Edwards, 1985; Smeral, Witt, & Witt, 1992), the most commonly used functional form in tourism demand analysis is the power model (Witt & Witt, 1995). Some research has found that the power model generally outperforms the linear model and fits the data better in terms of expected coefficient signs and statistical significance of the coefficients (Lee, Var, & Blaine, 1996; Witt & Witt, 1992; Vanegas & Croes, 2000). On top of those benefits, reasons for the power function’s popularity include (a) the more realistic value of the changing marginal relationship, (b) the simplification made possible by transforming the power model into a linear relationship using logarithms for estimation, and (c) the availability of demand elasticities through the estimated coefficients in transformed equations (Song et al., 2009). Particularly because of the third reason listed above, the double-log models (“log” on both sides of the equation) have been the predominant functional form in the group of power functions, including the log-linear and the linear-log models, to model the tourism demand over the past few decades. Consequently, in this study, we use the double-log function form, where the magnitude of the estimated income elasticity of demand can provide useful information for policy makers and tourism planners. That the ratio of tourist departure increases, decreases, or remains the same is the result of a change in per capita GDP and depends on the value of the income elasticity.

Econometric Specification

To estimate the impact of vacation policy changes on domestic tourism in China, we treat the dynamic structure of holiday propensity by considering preference changes as endogenous (we do this by including previous ratios of tourist departures in the model). The coefficient for the lagged dependent variable may be considered a measure of habit persistence and interdependent preferences (Garín-Muñoz, 2007). Except for the inclusion of previous dependent variables as an explanatory variable, the model presented in this study is a popularly used double-log tourism demand model in the sense that income, vacation policy, and travel cost are likely to play a central role in determining the ratio of urban residents’ departures. Ouerfelli (2008) indicated that, because of the diversity of economic and noneconomic factors that constitute the structure of the tourism sector, only the factors that can be measured would be shown in the analysis to make this process feasible. Therefore, from a theoretical point of view and with the demand elasticities considered, the demand for domestic tourism will be a double-log function of the quantity of tourism demand during the past decade, the consumer’s level of income, the average cost of a whole trip, the transport capacity of the originating cities, and several one-off events that occurred during the sample period. In addition, the interaction relationships among income, vacation policy, and location are also investigated in this study.

Thus, taking into account the consideration made in the preceding paragraphs, the ratio of tourist departures RQ_{i, t} from a city (i) during the year (t) would be a function of the quantity of the ratios of tourist departures during the last period, RQ_{i, t−1} the consumer’s level of income, per capita GDP_{i, t};³ the proxy of tourism price, TP_{i, t}; the transport capacity of originating cities, TC_{i, t} the policy dummy variable for controlling the change in vacation policy, D2007; and the event dummy variable for controlling the effects of SARS in 2003, D2003. Here, we lag RQ_{i, t} for 1 year: It is likely that past ratios of tourist departures have a promoting effect on current ratios. This can be explained by tourist expectations, habit persistence, “word-of-mouth,” path dependence, and constraints on supply. The resulting function would be

R Q_{i, t} = f (R Q_{i, t - 1}, G D P_{i, t}, T P_{i, t}, T C_{i, t}, D 2007_{t}, D 2003_{t})

From the generic function relationship, we examine three modeling specifications in order to stress the effects of vacation policy on Chinese domestic tourism demand and the interactive relationships between income, vacation policy, and transport capacity.

Model 1: Model of Demand Without Policy Variable

Because of the advantages of double-log functional forms stated above, the following specification is used as a basic model for estimating the determinants of domestic tourism demand in China:

\begin{array}{l} \ln R Q_{i, t} = β_{0} + β_{1} \ln R Q_{i, t - 1} + β_{2} \ln G D P_{i, t} + β_{3} \ln T P_{i, t} + β_{4} \ln T C_{i, t} + β_{5} D 2003_{t} + ε_{t} \\ i = 1, ..., N; t = 1, ..., T \end{array}

Here, except for dummy variables, both dependent variables and explanatory variables are transformed into logarithm forms. Thus, in Model 1, the ratio of tourist departures for Chinese urban residents depends on the number of tourist departures in the previous period, the per capita GDP, the transport capacities of originating cities, and the tourism price and special factors such as SARS in 2003.

Model 2: Model of Demand With Policy Variable but Without Interaction Relations

In Model 2, the policy variable D2007_t is included to capture the possible effects of policy changes of the Golden Weeks system on Chinese domestic tourism before and after 2007. Accordingly, the specification for this proposal is as follows:

\begin{array}{l} \ln R Q_{i, t} = β_{0} + β_{1} \ln R Q_{i, t - 1} + β_{2} \ln G D P_{i, t} + β_{3} \ln T P_{i, t} + β_{4} \ln T C_{i, t} + β_{5} D 2003_{t} \\ + β_{6} D 2007_{t} + ε_{t} \\ i = 1, ..., N; t = 1, ..., T \end{array}

Model 3: Model of Demand With Policy Variable and Interaction Relations

The third model measures the different effects of vacation policy on various income levels and various transport capacities. These effects can be represented by the interaction terms. Therefore, Model 3 is specified as follows:

\begin{array}{l} \ln R Q_{i, t} = β_{0} + β_{1} \ln R Q_{i, t - 1} + β_{2} \ln G D P_{i, t} + β_{3} \ln T P_{i, t} + β_{4} \ln T C_{i, t} + β_{5} D 2003_{t} \\ + β_{6} D 2007_{t} + β_{7} \ln G D P {}_{i, t}\times D 2007_{t} + β_{8} \ln T P_{i, t} \times D 2007_{t} + ε_{t} \\ i = 1, ..., N; t = 1, ..., T \end{array}

Methodological Issues

Estimation Method for Panel Data

To choose the estimation method for panel data, that is, the choice among pooled OLS, fixed-effects (FE) or random-effects (RE) model, the F test is first employed to check the homogeneity of city effects. Upon the result that city effects are found to exist, then the standard Hausman specification test is employed to help choose between the FE estimation or the RE estimation (Hausman, 1978). According to the Hausman test, the time-invariant unobserved variables are treated as part of the disturbances in the RE model and are assumed that their correlation with regressors is zero (Frondel & Vance, 2010). If the assumption is met, the RE estimators will be reported due to the advantages of greater efficiency over the FE estimators. Violation of the assumption, however, implies that the RE estimators are biased and only the FE estimation can produce consistent estimates.

On the basis of the empirical results presented in Table 3, the null hypothesis of the homogenous city effects for each of the variables is strongly rejected at a wide margin. This evidence suggests that OLS estimators are inefficient and may yield biased estimates. In addition, all results of Hausman test for selecting between FE or RE estimation reject the assumption of the RE model. Therefore, to save space, the FE estimation is only used here.

Table 3

F Test and Hausman Test for the Three Models

	Model 1	Model 2	Model 3
F test	0.0019**	0.0018**	0.0021**
Hausman test	0.0000***	0.0000***	0.0000***

p < .01. ***p < .001.

Thus, the error term in our models is composed of three parts: µ _it is assumed to be the remainder stochastic disturbance term, which is serially uncorrelated with zero mean and independently distributed across different originating cities in China.⁴ Meanwhile, µ_it is assumed to be uncorrelated with the initial condition RQ_{i, t} for t = 2, . . ., T, and with individual effects α_i for any t, according to Garín-Muñoz (2007). Furthermore, α_i denotes the unobserved city effect or a city fixed effect, which presents all factors affecting the ratio of tourist departures from a city that do not change over time. Geographical features, such as the city’s location in China, are included in α_i. Many other factors may not be exactly constant, but they might be almost constant over a 10-year period. These might include certain demographic features of the population (age composition, gender ratio, education level) and resident preference and habit. Different cities may have their own methods for developing tourism, and the residents in these cities might have different attitudes toward traveling; these are typically slow to change. Finally, λ_t represents the unobservable time effect, being as it is a component that remains invariable across a number of cities (Rodríguez et al., 2012). In the econometric model, the features of error terms, explanatory variables, and the relationship between them mainly determine the estimation procedure and estimators’ properties.

Then, Models 1, 2, and 3 are rewritten as follows:

Model 1: Model of demand without policy variable.

\begin{array}{l} \ln R Q_{i, t} = β_{0} + β_{1} \ln R Q_{i, t - 1} + β_{2} \ln G D P_{i, t} + β_{3} \ln T P_{i, t} + β_{4} \ln T C_{i, t} + β_{5} D 2003_{t} \\ + α_{i} + λ_{t} + μ_{i t} \\ i = 1, ..., N; t = 1, ..., T \end{array}

Model 2: Model of demand with policy variable but without interaction relations.

\begin{array}{l} \ln R Q_{i, t} = β_{0} + β_{1} \ln R Q_{i, t - 1} + β_{2} \ln G D P_{i, t} + β_{3} \ln T P_{i, t} + β_{4} \ln T C_{i, t} + β_{5} D 2003_{t} \\ + β_{6} D 2007_{t} + α_{i} + λ_{t} + μ_{i t} \\ i = 1, ..., N; t = 1, ..., T \end{array}

Model 3: Model of demand with policy variable and interaction relations.

\begin{array}{l} \ln R Q_{i, t} = β_{0} + β_{1} \ln R Q_{i, t - 1} + β_{2} \ln G D P_{i, t} + β_{3} \ln T P_{i, t} + β_{4} \ln T C_{i, t} + β_{5} D 2003_{t} \\ + β_{6} D 2007_{t} + β_{7} \ln G D P_{i, t} \times D 2007_{t} + β_{8} \ln T P_{i, t} \times D 2007_{t} + α_{i} + λ_{t} + μ_{i t} \\ i = 1, ..., N; t = 1, ..., T \end{array}

Panel Unit Root Test

The nonstationary of data is probably because of the presence of long run and seasonal unit roots, which imply that, in addition to a stochastic trend, these series exhibit a varying seasonality (Ouerfelli, 2008). To determine if a long-term co-integration relationship exists between vacation policy system and domestic tourism demand, it is necessary to test the panel unit roots. Because of the inclusion of the lagged dependent variable in this study, the Levin–Lin–Chu test is used to examine the panel data unit roots, testing the null hypothesis of nonstationary (Levin, Lin, & Chu, 2002). Test results and p value in Table 4 show that the variables in our models are stationary without unit roots, because the null hypothesis is rejected at the .01% significance level.

Table 4

Levin–Lin–Chu Test for Panel Unit Root

Coefficient	t Value	t Star	p > t
−0.7004	−9.300	−3.7480	.0001

Empirical Results and Discussion

Descriptive Data

Most pre-1990s tourism demand models made estimations using time series data without testing the stationary properties of the data and therefore have been criticized for generating spurious regression relationships (Song et al., 2009). Panel data makes it possible to control for unobserved omitted variables as long as those omitted variables do not change over time (Stock & Watson, 2011). The main advantage of panel data analysis over pure time series or pure cross-sectional modeling is that it features relatively large numbers of observations and a consequent increase in degrees of freedom, which reduces collinearity and improves the efficiency of the estimates (Song et al., 2009; Song et al., 2012). The availability of panel data in this study, therefore, will allow us to measure the effects of variables with few changes within cities and more variability across cities (Croes & Semrad, 2012; Garín-Muñoz & Montero-Martín, 2007). Annual Chinese domestic economic data in the 29 originating cities from 2001 to 2010 were used for this study, yielding a total of 290 observations obtained from CNTA and NBSC (2002-2011).

As illustrated in Table 5, the ratios of tourist departures of the 29 originating cities in 2001 and in 2010 clearly indicate that (a) the rapid growth of citizen domestic tourism took place in China during the past decade; (b) the common characteristics of the originating cities with the higher ratios in general are more affluent, enjoy better transportation infrastructure, and entertain higher travel proclivities; (c) all cities with more than 400% ratios, such as Dalian, Ningbo, Jinan, Guangzhou, and Xi’an, are endowed or in proximity to the well-known tourist attractions (namely the highest nationally ranked 4A grade tourist attractions). The accessibility and convenience to tourist attractions contribute to motivating higher tourist departures, thus promoting a fast-growing domestic tourist market; (d) meanwhile, the highest rates of growth are also found among these top five cities. It can be attributed that the municipal governments of these five cities almost invariably have tourism development as a priority, thus creating an enabling environment for citizen tourist departures and facilitating the highest rates of growth over the period of 10 years.

Table 5

Citizen Domestic Tourism in China

Ratio of Tourist Departures (%)
City	2001	2010	City	2001	2010
Beijing	171.9	198.3	Jinan	91.6	421.1
Tianjin	175.3	278.7	Zhengzhou	71.6	93.8
Huhehaote	70.9	93.1	Wuhan	91.8	381.1
Shenyang	124.2	267.0	Changsha	111.5	327.4
Dalian	83.5	460.4	Guangzhou	122.3	493.8
Changchun	69.9	117.7	Shenzhen	152.3	229.2
Harbin	76.5	193.7	Haikou	89.1	129.2
Shanghai	88.9	287.2	Chongqing	144.3	377.5
Nanjing	63.3	132.3	Chengdu	108.6	177.1
Wuxi	80.8	115.6	Kunming	164.5	184.0
Hangzhou	119.2	135.1	Xi’an	177.6	426.7
Ningbo	129.5	415.7	Lanzhou	85.5	312.4
Fuzhou	122.9	147.9	Xining	59.4	171.7
Xiamen	95.4	170.1	Wulumuqi	73.1	113.6
Nanchang	67.3	73.1

Source: China National Tourism Administration (1993-2011).

Note: The ratio of tourism departure is calculated by dividing the number of tourist departures by the resident population. As a result of residents in some cities taking two or more holidays in a year, the ratio of tourism departures in the Yearbook of China Tourism Statistics might be more than 100%.

Furthermore, Table 6 includes the descriptive statistics of all the variables used in the three models. It indicates that the mean and the standard deviation of the ratio of tourist departures are respectively, 148.46% and 88.44%, which shows the heterogeneity of the domestic tourist demand from different originating cities. The big standard deviation of the independent variable (per capita GDP_{i, t} undoubtedly shows the large difference in economic development and resident income level among these originating cities in China. It is noted that the enormous standard deviation of the independent variable (TC_i,t), which shows the significant difference in transportation infrastructure among different cities. It reveals that the transport capacity of some Chinese originating cities has been the constraining factor for domestic tourism development, particularly in the Golden Week holidays (Cai, 2009; Vanhove, 2011; Wu et al., 2012).

Table 6

Variables Used in the Three Models

Variables	Units	Range (2001-2010)	M	SD
RQ_i,t	Percentage	17.57-493.78	148.455	88.435
GDP_i,t	1 Yuan	5257-152099.2	36811.26	22035.64
TP_i,t	Percentage	91.6-128.3	101.427	3.713
TC_i,t	10,000 number	2141-156407.4	19801.660	23988.990

Note: RQ_i,t = The ratio of tourist departures from a city (i) during the year (t); GDP_i,t = The consumer’s level of income, per capita; TP_i,t = The proxy of tourism price; TC_i,t = The transport capacity of originating cities.

Estimation Results and Discussion

To estimate the three models above, annual data are used to avoid the problems associated with seasonality (Rodríguez et al., 2012). Taking into account the statistical information available and the dynamic characteristics of models, this study implements the generalized method of moments (GMM) procedure of Arellano and Bond (1991) for estimation in order to avoid the biased and inconsistent OLS estimators, within groups (WG) estimators (least squares after transformation to deviations from means); and random effects estimators. Several studies have demonstrated that the DIFF–GMM estimator proposed by Arellano and Bond (1991) was especially suitable for estimation in dynamic panel data models in which the number of economic units clearly exceeded those of time ranges (Garín-Muñoz, 2007; Kuo et al., 2008; Rodríguez et al., 2012; Wang, 2009), and this is the case in our sample.

Eliminating the individual effects by first differences and using the values of the dependent variable lagged two or more periods as valid instrument variables (IV) for the lagged dependent variable, the DIFF-GMM can generate consistent and efficient estimates of the parameters of interest in previous studies on tourism demand (Baltagi, 2008; Garín-Muñoz, 2006, 2007, 2009; Garín-Muñoz & Montero-Martín, 2007; Kuo et al., 2008; Rodríguez et al., 2012). Therefore, the dynamic panel procedures are applied to measure the main impact of several factors, especially vacation policy, on Chinese domestic tourism demand. The STATA v.11.2 econometric software is used to obtain the results of the estimation of the three proposed models using the DIFF-GMM procedure of Arellano and Bond (1991).

The validity of instruments for the lagged values of the endogenous and exogenous variables in the regressions determines the consistent estimation. The empirical results are shown in Table 7, which exhibits that the models perform satisfactorily. Assuming that there is no second-order autocorrelation in the errors as Arellano and Bond (1991) indicate, the test for the previous autocorrelation and the Sargan test of overidentifying restrictions are conducted. The test results indicate that it is impossible to reject the null hypothesis that assumes the consistency of the estimates. Meanwhile, the Wald test gives support to the joint significance of the explanatory variables. Most of these explanatory variables are also significant individually at the significance level of 10%, except for the tourism price variable in Model 2.

Table 7

Estimation Results for the Dynamic Models

Variables	Model 1	Model 2	Model 3
ln RQ(−1)	0.215 (.099)*	0.241 (.053)*	0.250 (.045)*
ln GDP	0.115 (.025)*	0.131 (.023)*	0.117 (.052)*
ln TP	−2.183 (.067)*	−1.868 (.117)	−2.621 (.066)*
ln TC	0.282 (.023)*	0.350 (.011)*	0.339 (.011)*
D2003	−0.179 (.000)***	−0.176 (.000)***	−0.169 (.000)***
D2007ln GDP × D2007		−0.068 (.054)*	−4.265 (.020)* 0.075 (.072)*
ln TP × D2007			−1.025 (.577)
F test	2.08 (.0019)**	2.08 (.0018)**	2.06 (.0021)**
Hausman test for random or fixed effects	44.24 (.0000)***	44.31 (.0000)***	44.24 (.0000)***
Levin–Lin–Chu test for unit root	−0.7004 (.0001)***
R² in fixed-effects model	.6644	.6705	.6733
Autocorrelation 2	−1.46	−1.38	−1.31
Sargan test (df)	24.36 (31)	25.68 (30)	24.83 (28)
Wald test (df)	55.13 (5)	63.51 (6)	66.55 (8)
No. of observations	232	232	232
Long-run parameters
ln GDP	0.146	0.173	0.156
ln TP	−2.781	−2.461	−3.495
ln TC	0.359	0.461	0.452

Note: Dependent variable = logarithm of the ratio of tourist departures. TP = tourism price; GDP = gross domestic product measured in per capita terms; TC = tourism capacity; D2007 = policy dummy variable for controlling the change in vacation policy; D2003 = event dummy variable for controlling the effects of SARS in 2003.

p < .1. **p < .01. ***p < .001 (for p value in parentheses).

The three columns in Table 7 show the estimated results for Models 1, 2, and 3, respectively. Since a double-log function form is used, the estimated coefficients are direct elasticities. The signs of the coefficients look theoretically reasonable as expected: a positive sign for the estimated coefficients of income (ln GDP) and transport capacity (ln TC), a negative sign for the estimated coefficients of tourism price (ln TP) and SARS dummy variable (D2003). As for other signs of variables we cannot predict before estimation, there is a positive sign for the lagged dependent variable, ln RQ(−1), and a negative sign for the policy variable, D2007. The positive ln RQ(−1) means that the ratio of tourist departures in previous years supplies the positive effects on the ratio of tourist departures in the current year. The negative D2007 means that the ratio of tourist departures decreased after vacation policy reform in China in 2007.

According to the results presented in Table 7, the positive and highly significant effect was obtained for the lagged dependent variable, indicating that habit persistence, “word of mouth” and path dependence are significant factors in explaining China’s domestic tourism demand. In fact, in all three models, about 25% of the ratio of tourist departures in China is attributed to the above effects. For example, the negative effects of the Golden Weeks policies were identified as especially related to travel deflation and the danger of overcrowding. The major implication of this result (the highly significant effects of the lagged dependent variable) for China’s domestic tourism market is that the early warning mechanism in popular scenery spots during Golden Weeks is crucial for leading and adjusting tourist flows. Moreover, inclusion of the lagged dependent variable could avoid overestimating the estimated coefficients for the rest of the variables, as Garín-Muñoz (2007) indicated. In order to further obtain the long-run elasticities, each of the estimated coefficients has to be divided by 1 − β₁. Thus, another advantage of using a dynamic model is the model’s ability to obtain both short-run elasticities and long-run elasticities (Garín-Muñoz, 2007). The last section of Table 7 reports the long-run elasticities of some revealing variables in this study.

The estimated coefficient for the income variable suggests that China’s domestic tourism demand has a weak causal relationship with the economic situation in different cities. A 1% decrease/increase in income leads to less than a 0.15% decrease/increase in the ratio of tourist departures. A low-income elasticity of demand implies that tourist departures are relatively insensitive to the level of economic development in the originating cities. The fact that the estimated value is less than 1 means that domestic trips have come within the reach of Chinese urban dwellers. Purchasing domestic tourism products is no longer regarded as a “luxury” privilege for the rich. This means that the Chinese domestic tourism market is becoming mature. However, it is worth noting that the interaction between the income term and policy variable is significant. It demonstrates that differences existed in the income effects on domestic tourism demand before and after the vacation policy reform in 2007. The facts indicate that the Chinese vacation policy reform, especially the public holiday arrangement, had effect on domestic tourism demand. In addition to money, whether having free time for travel has become a major determinant to affect tourism demand.

Transport capacity has a significant and positive impact on domestic demand in all of three models. This result shows that the rapid expansion of the transport capacity for each city may help improve the market of the tourism industry. The transport problems might not exist in some developed countries with less population. Faced with a huge population base in China, transport capacity becomes a dynamic driving force to domestic tourism demand as potential outflow of domestic tourists is hindered or delayed by limited transport capacity. Though the estimated elasticity for transport capacity is still less than 1, the increment of long-run elasticities indicates the importance of improving transport capacity.

We would also like to note that tourism price has the same negative effect on dependent variable in three models, though with different significant levels. In Model 3, a 1% decrease/increase in tourism price leads to about a 3% increase/decrease in the ratio of tourist departures. The price elasticity of demand indicates that tourist departures are relatively sensitive to the level of price rises in tourism industry. Because of the vast land size in China, the transportation cost to some domestic destinations might be more than that of outbound destinations. Based on studying domestic tourism demand, therefore, the price substitution effects are more obvious. In addition, some domestic tourism demands are also substituted by other leisure and recreational activities. Moreover, it is also worth noting that the interaction between the tourism price and policy variable is not significant. Wooldridge (2008) cautioned that, though sometimes interaction term is found statistically insignificant, its economic and practical significance cannot be ignored. In this case, the interaction term indicates economic or practical significance even though it has no statistical significance. This finding demonstrates that pricing for domestic tourism products was unaffected by the vacation reform policy in 2007 and the domestic tourism market did not experience significant pricing pressure after 2007, a possible sign of fierce competition by price in the domestic tourism market.

It is clear that vacation policy, the key variable we are interested in, has significant negative effects on domestic tourism demand. It is not difficult to imagine that the ratio of tourist departures would decrease following the adjustment of three Golden Weeks to two Golden Weeks and the institutionalization of five shorter public holidays. There are at least three reasons for this. The first reason was the direct effect of cutting short the May Day Golden Week. Though the total number of annual public holidays increased by 1 day with the new vacation policy in 2007, urban residents seem unwilling to use the three new shorter holidays to make their trips. These short holidays often need to move days from adjacent weekends to form a 3-day holiday. Thus, people have to occasionally work more than 1 week before or after the short holiday. More and more people start to substitute overnight travel with day-trip recreation activities or excursions (UNWTO definition for day trips) close to their residences, as evidenced by the sluggish ticket sales for long-distance train service for the 3-day May Day holidays in Guangzhou, but increasing demand for local taxi services as local residents opted to enjoy the spring in the suburbs (Sunshine Taxi, 2013). During the Dragon Boat Festival on June 10-12, 2013, marked increase of local residents touring their own cities was reported by online travel agencies (Zhejiang Business.com, 2013).⁵

Second, more than 10 years after the Golden Weeks system was introduced in China in the late 1990s, Chinese domestic tourists have become more prudent and practical in planning their travel schedules and consumptions. With more experiences and improved technology in travel planning, Chinese consumers have become more mature and rational in travel decisions. They pay more attention to the quality of experience, but not so much to travel quantity. Understanding the adverse outcomes from traveling during public holidays manifested in overcrowding, price hikes and declining service quality, many citizens prefer other recreation activities over travel. Parties with friends, visiting relatives, watching movies, recreational and fitness exercises have begun to replace travel during the shorter public holidays in China (Chinanews.com, 2013; People.com, 2013).

Finally, with the rapid economic development and opening policy in China, much domestic tourism demand has been substituted by the booming outbound tourism market (Figure 2). The Golden Weeks have become important windows for an increasing number of Chinese outbound tourists to travel to destinations all around the world (Gao, 2006). After the vacation policy reform in 2007, two Golden Weeks remained for travel consumption. In order to avoid overcrowding and diminishing travel experience at domestic destinations, increasing number of Chinese tourists prefer traveling overseas, especially to neighboring countries of Korean, Japan, Singapore, Malaysia, Thailand, and Vietnam, including the ever-popular cruise travel in east and southeast Asia.

Figure 2

China Outbound Tourism (million) 1992-2010

It is also important to note that the significant and negative effects of the SARS crisis in 2003 have been corroborated by the results. However, it is clear that the amended vacation policy in 2007 did not accomplish effectively its original intention of continuous growth while redistributing domestic demand evenly throughout the year by adding shorter public holidays because most Chinese people opt for leisure activities at home or close to home during the shorter public holidays, rather than visiting other cities. We also noted that the fixed effects are highly significant (p < .001) in all three models, indicating maturing domestic tourism market with physically and culturally endowed assets, diversified tourism products, coordinated regional development, and increasingly rational and loyal consumers. Obviously, endowed with rich and highly diversified natural, cultural and man-made tourism resources, China has great potential to meet increasingly differentiated domestic tourism demands.

Conclusion and Policy Implications

China has witnessed significant changes in its vacation policy system in the past decade. By establishing and estimating a dynamic panel data model based on China’s domestic tourism demand in the past 10 years, this study examines the impact of China’s 2007 vacation policy reform on domestic tourism demand using the econometric demand models as policy evaluations. Moreover, this study also contributes to the interpretation of domestic tourism demand theory in the Chinese context, demonstrating some unique Chinese characteristics under an existing theoretical framework.

Including the lagged dependent variable as a regressor, this study also uses the DIFF-GMM of Arellano and Bond (1991) for estimation. The estimated elasticities were plausible in terms of their economic signs, magnitudes, and statistical significance. One of the main conclusions of the study is the significant value of the lagged dependent variable (0.250), which may be interpreted as habit persistence, customer preference, path dependence, and “word-of-mouth” effects on consumer decision making in originating cities. Similar values have been obtained in previous studies in the case of tourism demand modeling (Garín-Muñoz, 2007; Garín-Muñoz & Montero-Martín, 2007; Kuo et al., 2008; Rodríguez et al., 2012; Wang, 2009).

The policy implication of this result is that relevant tourism organizations may take advantage of the consistency of tourist consumption habits and behaviors to anticipate, lead, and adjust tourist flow. This allows tourism organizations to improve the early warning mechanism that can help avoid travel deflation and the potential negative tourism experience associated with overcrowding. The dynamic model used also provides short- and long-run elasticities for the variables of interest, which is indicated as an additional advantage over most studies of static models by Garín-Muñoz (2007).

Additionally, this study reveals the causal relationship between Chinese domestic demand and special factors, such as income, tourism price, transport capacity, and vacation policy. The estimated values of the income elasticities suggest that the economic conditions of originating cities are still an important factor in determining domestic tourism demand in China, though both the short- and long-run income elasticities are less than 1. So, it would be very advisable to expand the potential tourist market in high-income areas, including Beijing, Shanghai, Guangzhou, and Shenzhen. Furthermore, the new types of tourism industry and leisure products, including spa, yacht, skiing, and golf should be vigorously developed to attract affluent Chinese consumers to domestic destinations.

The factor of tourism price to domestic tourism demand cannot be ignored. This study finds that Chinese domestic tourists are still sensitive to price. Though the ratios of tourist departures become higher in China, most of Chinese domestic tourist expenditure per trip remains lower. Differential pricing and marketing can be used by tourism enterprises to promote domestic tourists.

The domestic tourism market is also affected by the transport capacities of departure cities. This study implies that improving its transport system is a key to developing the tourism industry in China. Because of the unbalanced economic development in different parts of China, most potential tourism demand is concentrated in the big cities, which exerts more pressure to these cities’ transportation system. Breaking the traffic bottleneck is still a critical challenge that needs to be addressed for developing domestic tourism industry.

Finally, the significant and negative effect of the vacation policy variable is attributed to day-trip substitution effects, outbound travel substitution effects, and the maturity of tourist consumption. With the implementation of the reformed vacation policy in China since 2007, there has been intensified debates in recent years on the appropriate arrangement of the short public holidays and paid vacation implementation. This discourse focuses primarily on finding an effective system to arrange the public holidays. Currently, to form a 3-day short public holiday to coincide with the traditional Chinese festival, which often falls in the middle of the week, the 1-day public holiday has to be pieced together with a 2-day weekend. As a result, people need to work successively for 6 or 7 days before or after a 3-day holiday. Surveys show that more than half of the Chinese disapprove such public holiday arrangement. Unlike established and mature vacation system in developed countries, the current vacation policy needs further review and refinement in two critical aspects. First, the arrangement for the 3-day short public holiday could be structured by moving the 1-day public holiday to the weekend rather than shifting the 2-day weekend to the middle of the week, thus institutionalizing a permanent 3-day long weekend system observed by many developed countries. Second, the Ordinance on Paid Vacations for Employees must be implemented to encourage flexible travel planning by employees throughout the year, thus mitigating some adverse effects of temporal concentration of domestic travel during the Golden Weeks.

On a practical note, this study can serve as a reference for any stakeholders interested in China’s vacation policy reform and domestic tourism demand market, enabling them to gain a better appreciation of the relative factors that affect market scale. A better understanding of the relationship between China’s vacation policy system and consumer decision patterns will be of special value to policy makers in their efforts to boost the domestic and outbound markets.

Limitations and Future Research

A possible way of improving the results of this study would be to find the suitable proxy variable for tourism price in the models. Although it may be much easier to do research based on the perspective of destination with one or a few certain originating areas, it is difficult to collect data from originating areas. Therefore, because of the important role of tourism price in the tourist demand models, Instrumental Variable (IV) Estimation could be used as an alternative econometrics method for our models.

The methodology used in this study can be applied to other similar tourism demand studies. When data availability allows, it would be reasonable to use tourism earnings as a dependent variable to establish models for China’s domestic demand. Meanwhile, an obvious extension of this work could be divided into three subsamples according to different tourist motivations, such as sightseeing, visiting friends and relatives, or business and meetings. After estimating the effects of various factors on different market segments, the policy implications may be understood more practically.

Footnotes

Authors’ Note:

The article is sponsored by the China National Social Science Fund. The project number is 10ZD&051; and is also sponsored by the Beijing Talent Teacher Program. The insightful comments and suggestions of the three anonymous reviewers are gratefully acknowledged.

Notes

Chao Zhang, PhD (email: zhangchao@bisu.edu.cn), is an associate professor at the School of Hospitality Management, Beijing International Studies University in Beijing, China. Her research interests focus on tourist behavior and tourism economics. A.J. Singh, PhD, ISHC (email: singharj@msu.edu) is a professor in The School of Hospitality Business at Michigan State University in East Lansing, Michigan, USA. Larry Yu, PhD (email: lyu@gwu.edu), is a professor of management and tourism studies at the School of Business, the George Washington University in Washington DC.

References

Akis

(1998). A compact econometric model of tourism demand for Turkey. Tourism Management, 19, 99-102.

Arellano

Bond

(1991). Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. Review of Economic Studies, 58, 277-297.

Artus

J. R.

(1972). An economic analysis of international travel. IMF Staff Papers, 19, 579-614.

Baltagi

(2008). Econometric analysis of panel data (4th ed.). Chichester, England: Wiley.

Barry

O’Hagan

(1972). An econometric study of British tourist expenditure in Ireland. Economic and Social Review, 3, 143-161.

Burkart

A. J.

Medlik

(1981). Tourism: Past, present and future (2nd ed.). Oxford, England: Heinemann.

Cai

J. M.

(2009). A political economics analysis of the China’s holiday system reform. Study and Exploration, 5, 2-12.

Chen

S. Y.

Y. P.

(2007). An empirical analysis of the golden week holiday system and model for reform of vacation policies. Consumer Economics, 23(5), 44-50.

China National Tourism Administration. (1993-2011). The yearbook of China tourism statistics. Beijing: China Tourism Press.

10.

Chinanews.com. (2013). The shorter May Day holidays usher the fitness “spurt” of Shanghai residents. Retrieved from http://www.chinanews.com/ty/2013/04-29/4776936.shtml

11.

Cooper

Scott

Kester

(2006). New and emerging markets. In Buhalis

Costa

(Eds.), Tourism business frontiers: Consumers, products and industry (pp. 19-29). London, England: Elsevier.

12.

Croes

Semrad

K. J.

(2012). Does discounting work in the lodging industry? Journal of Travel Research, 20(10), 1-15.

13.

Croes

Vanegas

Sr (2005). An econometric study of tourist arrivals in Aruba and its implications. Tourism Management, 26, 879-890.

14.

Crouch

G. I.

Oppewal

Huybers

Dolnicar

Louviere

Devinney

(2007). Discretionary expenditure and tourism consumption: Insights from a choice experiment. Journal of Travel Research, 45, 247-258.

15.

Crouch

G. I.

Schultz

Valerio

(1992). Marketing international tourism to Australia: A regression analysis. Tourism Management, 13, 196-208.

16.

Di Matteo

(1993). The determinants of expenditures by Canadian visitors to the United States. Journal of Travel Research, 31, 34-42.

17.

Dritsakis

Athanasiadis

(2000). An econometric model of tourist demand: The case of Greece. Journal of Hospitality & Leisure Marketing, 7, 39-49.

18.

Edwards

(1985). International tourism forecasts to 1995: EIU special report No. 188. London, England: Economist.

19.

Esping-Anderson

(1990). The three worlds of welfare capitalism. Cambridge, England: Polity Press.

20.

Eugenio-Martin

Sinclair

M. T.

Yeoman

(2005). Quantifying the effects of tourism crises: An application to Scotland. Journal of Travel & Tourism Marketing, 19, 21-34.

21.

Fayos-Sola

Bueno

A. P.

(2001). Globalization, national tourism policy and international organizations. In Cooper

C. P.

Wahab

(Eds.), Tourism in the age of globalization (pp. 195-210). London, England: Routledge.

22.

Font

A. R.

(2000). Mass tourism and the demand for protected natural areas: A travel cost approach. Journal of Environmental Economics and Management, 39, 97-116.

23.

Frondel

Vance

(2010). Fixed, random, or something in between? A variant of Hausman’s specification test for panel data estimators. Economics Letters, 107, 327-329.

24.

Gao

S. L.

(2006). Studies on policies of the tourism industry in China. Beijing: China Tourism.

25.

Gao

(2007). A historical account of vacation policies in the world. Government Legal System, 9(2), 32.

26.

Garín-Muñoz

(2006). Inbound international tourism to Canary islands: A dynamic panel data model. Tourism Management, 27, 281-291.

27.

Garín-Muñoz

(2007). German demand for tourism in Spain. Tourism Management, 28, 12-22.

28.

Garín-Muñoz

(2009). Tourism in Galicia: Domestic and foreign demand. Tourism Economics, 15, 753-769.

29.

Garín-Muñoz

Montero-Martín

(2007). Tourism in the Balearic Islands: A dynamic model for international demand using panel data. Tourism Management, 28, 1224-1235.

30.

Garín-Muñoz

Pérez-Amaral

(2000). An econometric model for international tourism flows to Spain. Applied Econometrics Letters, 7, 525-529.

31.

Gray

(1966). The demand for international travel by the United States and Canada. International Economic Review, 7, 83-92.

32.

H. M.

C. L.

(2003). The impact of the evolution of income distribution structure in China on the domestic tourism consumption. Tourism Tribune, 18(2), 19-23.

33.

Han

Durbarry

Sinclair

M. T.

(2006). Modeling US tourism demand for European destinations. Tourism Management, 27, 1-10.

34.

Hausman

J. A.

(1978). Specification tests in econometrics. Econometrica, 46, 1251-1271.

35.

Ismail

J. A.

Iverson

T. J.

Cai

L. A.

(2000). Forecasting Japanese arrivals to Guam: An empirical model. Journal of Hospitality & Leisure Marketing, 7, 51-63.

36.

Jensen

T. C.

(1998). Income and price elasticities by nationality for tourists in Denmark. Tourism Economics, 4, 101-130.

37.

Kim

Song

H. Y.

(1998). An empirical analysis of demand for Korean tourism: A co-integration and error correction approach. Tourism Analysis, 3, 25-41.

38.

Kliman

M. L.

(1981). A quantitative analysis of Canadian overseas tourism. Transportation Research, 15, 487-497.

39.

Kuo

H. I.

Chen

C. C.

Tseng

W. C.

L. F.

Huang

B. W.

(2008). Assessing impacts of SARS and Avian Flu on international tourism demand to Asia. Tourism Management, 29, 917-928.

40.

Kwack

S. Y.

(1972). Effects of income and prices on travel spending abroad 1960III-1976IV. International Economics Review, 13, 245-256.

41.

Lee

C. K.

Var

Blaine

T. W.

(1996). Determinants of inbound tourism expenditure. Annals of Tourism Research, 23, 527-542.

42.

Levin

Lin

C. F.

Chu

(2002). Unit root tests in panel data: Asymptotic and finite sample properties. Journal of Econometrics, 108, 1-24.

43.

Lew

Zhang

(2003). Tourism in China. New York, NY: Haworth Press.

44.

Song

H. Y.

Witt

S. F.

(2006). Forecasting tourism demand using econometric models. In Buhalis

Costa

(Eds.), Tourism management dynamics: Trends, management and tools (pp. 219-228). Oxford, England: Elsevier.

45.

Y. W.

Xia

L. G.

(2004). An analysis of tourism consuming structure of China’s urban residents. Tourism Science, 18(2), 30-38.

46.

Lim

(1997). Review of international tourism demand models. Annals of Tourism Research, 24, 835-849.

47.

Lim

(2004). The major determinants of Korean outbound travel to Australia. Mathematics and Computers in Simulation, 64, 477-485.

48.

Little

J. S.

(1980). International travel in the US balance of payment. New England Economic Review, May/June, 42-55.

49.

Loeb

(1982). International travel to the United States: an economic evaluation. Annals of Tourism Research, 9, 7-20.

50.

Martin

C. A.

Witt

S. F.

(1988). Substitute prices in models of tourism demand. Annals of Tourism Demand, 15, 255-268.

51.

Morley

C. L.

(1994). The use of CPI for tourism prices in demand modeling. Tourism Management, 15, 342-346.

52.

Morley

C. L.

(1997). An evaluation of the use of ordinary least squares for estimating tourism demand model. Journal of Travel Research, 35, 69-73.

53.

Morley

C. L.

(1998). A dynamic international demand model. Annals of Tourism Research, 25, 70-84.

54.

National Bureau of Statistics of China. (2002-2011). China statistical yearbook (for various years). Beijing: China Statistics Press.

55.

O’Rourke

(2003). Outsourcing regulation: Analyzing non-governmental systems of labor standards and monitoring. Policy Studies Journal, 31, 1-29.

56.

Ouerfelli

(2008). Co-integration analysis of quarterly European tourism demand in Tunisia. Tourism Management, 29, 127-137.

57.

People.com. (2013). All scenic attractions in Chuzhou embraced waves of tourism during the Qingming shorter holiday. Retrieved from http://tour.jschina.com.cn/system/2013/04/08/016828537.shtml

58.

Perry

(2004). A look inside America: Exploring America’s cultural values and holidays. Philadelphia, PA: Multi-Language Media.

59.

H. L.

Lam

(1997). A travel demand model for Mainland Chinese tourists to Hong Kong. Tourism Management, 18, 593-597.

60.

Ren

(2007). The analysis of gains and losses for travel in the Golden Week. Journal of Mianyang Normal University, 26(7), 18-20.

61.

Rodríguez

X. A.

Martínez-Roget

Pawlowska

(2012). Academic tourism demand in Galicia, Spain. Tourism Management, 33, 1583-1590.

62.

Schor

(1991). The overworked American: The unexpected decline of leisure. New York, NY: Basic Books.

63.

Shao

Q. W.

(2008). Speech on the national tourism work meeting 2008. CNTA Web News. Retrieved from http://www.cnta.gov.cn

64.

Smeral

Witt

S. F.

Witt

C. A.

(1992). Econometric forecasts: Tourism trends to 2000. Annals of Tourism Research, 19, 450-466.

65.

Song

H. Y.

Dwyer

Cao

(2012). Tourism economics research: A review and assessment. Annals of Tourism Research, 39, 1653-1682.

66.

Song

H. Y.

Lin

(2010). Impacts of the financial and economic crisis on tourism in Asia. Journal of Travel Research, 49, 31-38.

67.

Song

H. Y.

Witt

S. F.

(2003). Tourism forecasting: The general-to-specific approach. Journal of Travel Research, 42, 65-74.

68.

Song

H. Y.

Witt

S. F.

(2009). The advanced econometrics of tourism demand. London, England: Routledge.

69.

Song

H. Y.

Wong

K. K. F.

(2003). Tourism demand modeling: A time-varying parameter approach. Journal of Travel Research, 42, 57-64.

70.

Stock

J. H.

Watson

M. W.

(2011). Introduction to econometrics (3rd ed., pp. 318). New York, NY: Wiley.

71.

Summary

(1987). Estimation of tourism demand by multivariable regression analysis: Evidence from Kenya. Tourism Management, 12, 317-322.

72.

Sun

Y. X.

(2009). The macro analysis of tourist consumption structure and contribution in Shanghai. East China Economic Management, 23(12), 1-3.

73.

Sunshine Taxi. (2013). Sluggish train ticket sales, but great demand for taxi services during the May Day shorter holidays. Retrieved from http://gzzcw.cn/html/2013/4/20/14717.html

74.

Uysal

Crompton

J. L.

(1985). An overview of approaches used to forecast tourism demand. Journal of Travel Research, 24, 7-15.

75.

Vanegas

Sr. Croes

R. R.

(2000). Evaluation of demand: US tourists to Aruba. Annals of Tourism Research, 27, 946-963.

76.

Vanhove

(2011). The economics of tourism destinations. London, England: Elsevier.

77.

Var

Mohammad

Icoz

(1990). Factors affecting international tourism demand for Turkey. Annals of Tourism Research, 17, 606-610.

78.

Wang

X. G.

Zhang

J. H.

Ren

L. X.

Liu

Z. H.

Zhang

(2012). Investigating influence factors of tourism consumption coupon effect based on structural equation model approach: A case study of Hangzhou. Geographical Research, 31, 543-554.

79.

Wang

Y. S.

(2009). The impact of crisis events and macroeconomic activity on Taiwan’s international inbound tourism demand. Tourism Management, 30, 75-82.

80.

Wen

Tisdell

C. A.

(2001). Tourism and China’s development. Singapore: World Scientific.

81.

Wick

(2001). Workers’ tool or PR ploy? A guide to codes of international labor practice. Bonn, Germany: Friedrich-Ebert-Stiftung.

82.

Witt

S. F.

(1980a). An abstract mode—Abstract (destination) node model of foreign holiday demand. Applied Economics, 12, 163-180.

83.

Witt

S. F.

(1980b). An econometric comparison of UK and German foreign holiday behavior. Managerial and Decision Economics, 1, 123-131.

84.

Witt

S. F.

Martin

C. A.

(1987). International tourism demand models: Inclusion of marketing variables. Tourism Management, 8, 33-40.

85.

Witt

S. F.

Song

H. Y.

Louvieris

(2003). Statistical testing in forecasting model selection. Journal of Travel Research, 42, 151-158.

86.

Witt

S. F.

Witt

C. A.

(1992). Modeling and forecasting demand in tourism. London, England: Academic Press.

87.

Witt

S. F.

Witt

C. A.

(1995). Forecasting tourism demand: A review of empirical research. International Journal of Forecasting, 11, 447-475.

88.

Wong

K. K. F.

Song

H. Y.

Witt

S. F.

D. C.

(2007). Tourism forecasting: To combine or not to combine? Tourism Management, 28, 1068-1078.

89.

Wooldridge

J. M.

(2008). Introductory econometrics: A modern approach. Chicago, IL: South-Western, Division of Thomson Learning.

90.

B. H.

Xue

Morrison

A. M.

Leung

X. Y.

(2012). Frame analysis on golden week policy reform in China. Annals of Tourism Research, 39, 842-862.

91.

Yang

(2005).The rational return of tourism during the Golden Week holidays. Tourism Science, 2, 49-51.

92.

York

Q. Y.

Zhang

H. Q.

(2010). The determinants of the 1999 and 2007 Chinese golden holiday system: A content analysis of official documentation. Tourism Management, 31, 881-890.

93.

Zhang

(2008). The impact of Golden Week on Chinese tourism. Commercial Research, 3, 172-177.

94.

Zhang

(2003). China’s tourism since 1978: Policies, experiences, and lessons learned. In Lew

Zhang

(Eds.), Tourism in China (pp. 13-34). New York, NY: Haworth Press.

95.

Zhejiang Business.com. (2013). Picking fruits and water tours are the most popular activities around Hangzhou during the Dragon Boat Festivalholiday. Retrieved from http://biz.zjol.com.cn/05biz/system/2013/06/06/019388654.shtml