Tourism development and regional production efficiency: Evidence from southwestern China

Abstract

This study examines the relationship between the production efficiency of a region and tourism flows. We use stochastic frontier production models with spatial effects based on data from 46 regions in southwestern China, most of which are economically underdeveloped. The results show that tourist flows (TFs) into a region are significantly related to the region’s production efficiency. The TFs in the neighboring regions also correlate with the production efficiency of a region. This study validates the tourism-led efficiency enhancement proposition at the regional level and has rich implications for practice.

Keywords

regional production efficiency regional tourism stochastic frontier production function tourism-led efficiency enhancement

Introduction

Researchers and many other experts consider tourism development an effective way for regions to improve their local economies. Many studies support the tourism-led growth hypothesis, including those by Balaguer and Cantavella-Jorda (2002), Brida and Risso (2009), and Figini and Vici (2010). The tourism-led hypothesis suggests a causal relationship between tourism and the growth of regional economies. According to the literature, the primary mechanisms that explain how tourism helps regional economies include employment creation (Durbarry, 2002), revenue creation (Durbarry, 2004; Sinclair, 1998), and the multiplier effect (Singh, 2008). Marrocu and Paci (2011), however, regard tourist flows (TFs) in a region as an important source of knowledge and information, proposing that local businesses can extract valuable intellectual input from tourists. Consequently, regional corporate innovation may increase and regional production efficiency may improve. Marrocu and Paci’s (2011) proposition is referred to as tourism-led efficiency enhancement (TLE).

This proposition deserves attention from researchers because it has broad implications for practice. Following Marrocu and Paci (2011), no research explores this issue further. To address this deficit, this study investigates the association between tourism and regional production efficiency using economically underdeveloped regions as the analysis objects. The study uses 46 cities (or 46 regions) in southwestern China as research samples, the majority of which are economically poor.

This study contributes to the literature as follows: first, it is the first empirical work utilizing the panel data of underdeveloped regions to demonstrate that tourism plays a significantly positive role in the enhancement of regional production efficiency. For underdeveloped regions, the tourism-led efficiency proposition is particularly relevant because these regions generally lack alternative ways, such as intensive research and development (R&D) activity, training, education, and foreign direct investment (FDI), to lift their production efficiency.

Second, this study demonstrates the spatial effect of tourism reflected in regional production efficiency, which suggests that TFs to neighboring regions result in production-efficiency enhancement for a given region. This type of spatial effect refers to a new rationale for interregional tourism cooperation.

Third, this study has merit in its research methodology. To test the tourism-led efficiency proposition, Marrocu and Paci (2011) use total factor productivity to measure the production efficiency of a region, an approach that relies on an unrealistic assumption that the production of a region is always conducted in the most effective way. In contrast to Marrocu and Paci (2011), this study adopts the stochastic frontier production approach, a method commonly used for measuring production efficiency (Filippini et al., 2008). In the stochastic frontier production approach, the gap between the current level and the frontier level of production serves as an indicator of inefficiency, and regional tourism is included as one variable to determine regional inefficiency scores whereby the impact of tourism is revealed. To the best of our knowledge, this study represents the first application of the stochastic frontier production approach in explaining the link between regional tourism and regional production efficiency.

Literature review

The link between tourism and regional productivity is rooted in trade theory because regional tourism is a type of trade across regions. Traditionally, the trade literature emphasizes that trade may result in many significant outcomes, such as production specialization (Young, 1991) and increased market size (Helpman and Krugman, 1985), which both help to boost regional productivity.

Recently, the trade literature has conducted analyses from the perspective of knowledge transfer, recognized as a critical factor in the endogenous growth of a modern economy (Marrocu and Paci, 2011; Navaretti and Tarr, 2000). In an economic system, trading products and service always entails knowledge transmission. Navaretti and Tarr (2000) emphasize that the productivity of a region is greater when trade provides access to more sophisticated technologies or knowledge. Many empirical studies validate trade as a channel of learning that leads to increased total factor productivity for a region, especially when an underdeveloped region trades with more developed regions with large stocks of knowledge (Kim et al., 2007).

For regional tourism, the following logic supports the tourist-led efficiency proposition.

Knowledge transfer by tourist/local interactions

As a result of regional tourism development, more tourists flow into a region, bringing valuable knowledge and information that may stimulate product innovation and productivity improvement in the region’s firms. The knowledge transfer that occurs in tourist interactions with locals becomes more valuable when a destination is less developed. Compared to local people in a poor region, arriving tourists often possess more advanced knowledge and information. Face-to-face communication with tourists or close observation of tourist behavior can be important ways for local people to understand better what is happening outside their area. As Navaretti and Tarr (2000) point out, there are alternative channels for local people to acquire knowledge and information, including schooling and education, intensive R&D activities, and FDI. But these alternative channels cannot function effectively in poor regions due to the lack of funding for schooling, education, and R&D activities, and the lack of needed conditions to attract FDI. Under these circumstances, the channel value of tourism becomes especially critical, as the following example from China clearly demonstrates.

WeChat Payment is a high-efficiency payment platform for businesses that saves on cash management expenses and enhances the operational efficiency of firms. Initially, WeChat Payment was widely used in restaurants, accommodations, taxis, and other service industries in eastern China, while business operators in western China (where underdeveloped regions predominate) had no knowledge of the system. When tourists traveled from eastern to western China, they frequently asked to settle bills via WeChat. Managers in western China thus gained awareness of WeChat Payment’s usefulness and gradually adopted it. As a result of adopting WeChat, the operational efficiency of service industries in western China has improved (Wu and Lee, 2017).

The externalities of tourism

The impact of tourism may extend across the actual boundaries of the tourism industry in a region. This cross-boundary effect constitutes the externalities of tourism.

For regional productivity, the first type of externality may be local firms’ access to the knowledge imported into a region by tourists. Much management literature highlights the role of contiguity between firms and customers for innovation. Von Hippel (1978) proposes a customer-active paradigm where firms collect ideas from customers to generate new industrial products. Foxall and Johnston (1987) also emphasize the importance of customer knowledge on product innovation. Similarly, Joshi and Sharma (2004) propose that access to customer knowledge, including customer preferences, promotes the success of new products. Marrocu and Paci (2011) state that tourists who visit a destination are generally customers or potential customers of local firms. These firms are able to increase their contacts with these customers at no additional cost while often improving production efficiency (Baek and Lee, 2018).

The second type of externality, which is related to regional production efficiency, may lie in the immigration and investment regional tourism stimulates. Feng and Page (2000) analyze the probability that tourists will make a decision to immigrate into a visited destination. Herein, we take Lijiang city, Yunnan province (one research sample in this study) as a case example. A few decades ago, Lijiang city initiated the development of tourism industry. As the tourism industry boomed, many visitors from developed regions in China (including Zhejiang and Guangdong provinces) chose to immigrate to Lijiang city to run various types of businesses and live as Lijiang residents. The valuable knowledge and ideas these new immigrants brought to the city served as resources that Lijiang city was able to exploit to innovate and improve production efficiency. Additionally, tourists may be potential investors for a destination (Fereidouni and Al-mulali, 2014) and their investments may have a positive impact on regional production efficiency.

The third type of externality relates to the infrastructure (especially for transportation) and capital (including human capital) a region acquires from tourism industry development. The tourism industry of a region may share the infrastructure built in connection with its development with other industries within the same region. What benefits tourists can concurrently benefit local firms at no additional cost. For example, as one critical piece of infrastructure, transport capacity may initially be constructed to ensure mobility for tourists, but the same capacity can later be leveraged by other regional industries, as when the local manufacturing sector takes advantage of improved roads to reach new markets. Many empirical studies demonstrate how transportation infrastructure significantly improves regional production efficiency (Aschauer, 1989; Bonaglia et al., 2000). Tourism development also supports the accumulation of human capital. By servicing the tourism industry, local people may improve their experience and competency in service, communication, team-collaboration skills, and commercial awareness, all of which can be transferred to other industries or sectors, contributing to productivity improvement outside the tourism sector (Kireyeva et al., 2018).

Given that knowledge or information is easily transportable, the positive effect of tourism on regional productivity is not limited to a certain geography range. A given region also benefits from tourism development in its neighboring regions in terms of productivity, establishing the spatial effect of tourism on regional productivity.

Although the literature supports the tourism-led efficiency proposition, it still seems reasonable to put forward an opposing proposition that regional tourism cannot contribute to regional production efficiency. The reasoning is: unlike the high-tech industry, the tourism industry is typically of low productivity and generally does not require a labor force with a high-level skill or advanced knowledge. When the economy of a region relies on the tourism industry, it is possible that its productivity may be locked at a low level (Lee and Syah, 2018).

Given this opposing proposition, we need to conduct empirical studies to test the relationship between tourism and regional productivity. Marrocu and Paci (2011) use data from 15 European Union countries to validate a positive relationship between tourism and regional productivity, but their research is incomplete since they did not take into account the spatial effect of tourism. Their conclusions are based solely on examining developed regions, whereas the tourism-led efficiency proposition is most appropriate for underdeveloped regions.

Research methods

Study area

Economic development at the regional level in China is largely unbalanced. Unlike eastern China (the most developed area in China) or China as a whole, southwestern China suffers from widespread poverty and has exhibited a low level of economic growth over a long term. Due to its harsh geographical environment and remoteness, southwestern China is isolated from the outside world for long stretches of time. Residents in this region have difficulty accessing knowledge or information from the outer world, which explains in large part why this area is economically backward. Figure 1 illustrates a huge gap between southwestern China and eastern China and China as a whole in terms of annual gross domestic product (GDP) over the research period.

Figure 1.

Comparison of regional economy development. East China, including Zhejiang province, Jiangsu province, Anhui province, and Shanghai city, represents the most economically developed region in China. The curve of GDP per capita of China indicates the average level of economic development in China over the research time period of this study. GDP: gross domestic product.

However, southwestern China boasts rich natural resources and a diverse cultural heritage, both of which may be exploited to attract tourists. Due to its remoteness, the natural environment of southwestern China has been well preserved and remains in its primitive condition. Its cultural heritage is rooted in the rich and unique culture of its local minority nationalities. There are nearly 30 minority nationalities living in different regions in southwestern China, including the Tibetan, Mongol, Li, Tujia, Jingpo, and Naxi peoples. Relying on such unique resource advantages, local governments throughout southwestern China frequently put tourism development among the top priorities for boosting regional economic activity. As a result, many regions in southwestern China have experienced a rapid expansion of their tourism industries, including Lijiang city in Yunnan province and Leshan city in Sichuan province.

We chose cities in southwestern China affiliated with Yunnan, Sichuan, and Guizhou provinces as our research samples. Chongqing municipality was historically a part of Sichuan province but has been administrated directly by the central government of China since 1997. We do not include cities in Chongqing municipality because they are highly industrialized. Chengdu, Kunming, and Guiyang (the capitals of Sichuan, Yunnan, and Guizhou, respectively) have highly developed economies. We include these capitals to ensure the geographical completeness of the research area, allowing us to estimate the spatial effects of tourism. In total, we have 46 cities as research samples, the locations of which are indicated in Figure 2.

Figure 2.

Study area and sample cities. We have 21 sample cities in Sichuan province, 16 sample cities in Yunan province, and 9 sample cities in Guizhou province. In total, we have 46 sample cities.

Table 1 reports the dynamic development of tourism for all sample cities, indicating a trend of continually rising tourist arrivals and tourism revenue (TR). Total tourist arrivals in all our sample cities in 2014 exceeded 1.3 billion persons, about eight times the total population in these regions. In 2014, in the research area, the contribution of TR to annual GDP was up to 19%, demonstrating that tourism plays a critical role for regional development in Southwest China.

Table 1.

Dynamics of tourist arrivals and TR in the research area.

	2005	2006	2007	2008	2009	2010	2011	2012	2013	2014
TF	236,053.68	301,134.07	352,911.10	376,656.13	460,387.33	553,672.41	716,480.10	853,931.55	1,570,411.65	1,316,758.214
Population	164,487.70	166,419.13	167,014.91	167,223.32	167,698.77	161,400.21	161,500.34	162,192.73	162,958.23	164,131.00
TF/population	1.44	1.81	2.11	2.25	2.75	3.43	4.44	5.26	9.64	8.02
TR	128,199.47	175,353.43	211,258.11	228,080.54	313,626.06	401,733.25	560,153.85	686,076.74	867,064.86	997,450.253
GDP	1,304,347.74	1,518,080.22	1,826,802.32	2,160,398.88	2,468,539.06	2,948,189.73	3,620,679.607	4,253,524.67	4,786,928.42	5,309,292.25
TR/GDP	0.10	0.12	0.12	0.11	0.13	0.14	0.15	0.16	0.18	0.19

Note: TF: tourist flow; TR: tourism revenue; GDP: gross domestic product. The unit of TF (tourist arrivals of sample cities) is one thousand person-times; the unit of population is one thousand persons; the unit of TR (TR of sample cities) and annual GDP is one million Yuan.

Econometric models

The stochastic frontier approach is one of the primary methods used for measuring efficiency, including production efficiency and cost efficiency at the firm or regional levels. The stochastic frontier approach holds an advantage over the data enveloping analysis approach (another method for assessing efficiency) since it isolates the influence of factors other than inefficient behaviors, correcting the possible upward bias of inefficiency from the deterministic model (Chen, 2007).

According to the stochastic frontier approach, the production efficiency of a firm (or region) reflects how well a firm (or region) processes inputs to achieve its outputs, compared to the level of maximum potential indicated by the frontier of production possibility (Bernini and Guizzardi, 2010; Wu, 2000). Aigner et al. (1977) initially proposed the stochastic frontier approach, and there are some studies in the tourism literature that use this approach to investigate efficiency issues in the tourism industry (Arbelo et al., 2017; Assaf et al., 2016; Bernini and Guizzardi, 2010; Chen, 2007; Oliveira et al., 2014).

In line with Wang and Schmidt (2002) and Bernini and Guizzardi (2010), we define a stochastic frontier production function as follows

ln y_{i t} = X_{i t} β + (ν_{i t} - u_{i t})

In equation (1), $y_{i t}$ represents the real GDP of region i at time t, a measure of regional output; $X_{i t}$ represents a set of log form inputs, typically as capital ( $K_{i t}$ ) and labor ( $L_{i t}$ ). As a benchmark, this study uses a trans-log production function as equation (2) to define how a region transforms inputs into outputs

ln y_{i t} = β_{0} + β_{k} ln K_{i t} + β_{l} ln L_{i t} + β_{k k} {(ln K_{i t})}^{2} + β_{k l} ln K_{i t} \times ln L_{i t} + β_{l l} {(ln L_{i t})}^{2} + (ν_{i t} - u_{i t})

Compared with a traditional Cobb–Douglas production function, a trans-log production function has many distinct advantages. In a trans-log production function, there is no a priori restriction regarding the internal scale return of capital and labor, and hence the marginal contributions of inputs are flexible. The other flexibility inherent in a trans-log production function lies in that the marginal contribution of one input partly depends on the level of another input.

$ν_{i t} - u_{i t}$ is the composite disturbance term, where $ν_{i t}$ is a purely random disturbance and $μ_{i t}$ indicates a nonnegative random disturbance indicating the production inefficiency of region i.

Accordingly, the following definitions are adopted

ν_{i t} : N (0, δ_{ν}^{2})

u_{i t} : N^{+} (μ_{i t}, δ_{u}^{2})

μ_{i t} = α z_{i t}

Equation (5) is an inefficiency function (Deprins and Simar, 1989), where $z_{i t}$ represents a set of variables which determine the inefficiency level of a region, and $α$ represents the coefficients of $z_{i t}$ .

The basic inefficiency function is defined as follows

μ_{i t} = α_{0} + α_{lnTF} lnTF + α_{lnFDI} lnFDI + α_{lnATC} lnATC + α_{lnEE} lnEE + α_{lnPI} lnPI + α_{lnPF} lnPF + ε_{i t}

In equation (6), TF ( ${TF}_{i t}$ ) is one of the main explanatory variables, indicated by tourist arrivals to region i at time. Other regional-specific variables should be controlled if they are theoretically efficiency determinants. FDI in a region reflects the link between sample regions and the outer world and supports innovation activity, and is positively related to regional efficiency (Cheung and Ping, 2004). According to Avila and Evenson (2010), the accumulated technological capital (ATC), represented by the number of accumulated patent applications, strongly relates to the efficiency performance of a region. The expenditure for education (EE) in a region, a proxy for the regional investment in education, may indicate the human capital of a region and may have a positive influence on regional productivity (Temple, 2001). We use the proportion of the industrial added value of the primary industry (PI) in regional GDP to reflect the industrial structure of a region, given that the PI (including industries such as agriculture, forestry, fishery, etc.) is generally related to low production efficiency. Additionally, we introduce the proportion of regional public finance (PF) (the regional government’s spending) in regional GDP, to control for the behavior paradigm of governments. According to Wang et al. (2006), greater PF usually refers to a conservative behavior paradigm by a government, which means that it has a strong incentive to protect local or state-owned firms since these firms usually are the main sources of revenue for a government. However, the production efficiencies of local or state-owned firms are generally low, decreasing the production efficiency of a region.

The extended inefficiency function, which takes into account the spatial effect of tourism, is defined as follows

μ_{i t} = α_{0} + α_{lnTF} lnTF + α_{W lnTF} W lnTF + α_{lnFDI} lnFDI + α_{lnATC} lnATC + α_{lnEE} lnEE + α_{lnPI} lnPI + α_{lnPF} lnPF + ε_{i t}

In equation (7), the spatially lagged dependent variable, $W ln {TF}_{i t}$ , is used to capture the spatial effect of tourism on regional efficiency. The spatial effect reflects that tourism activities happening in neighboring regions may lead to a reduction in production inefficiency of a particular region. $W ln {TF}_{i t} = \sum_{j = 1}^{N} w_{i j} {TF}_{j t}, j \neq i$ , where $W$ is a spatial weighting matrix and $w_{i j}$ is an element of the spatial weighting matrix that measures the spatial connection between regions i and j. We use a contiguity matrix (Marrocu and Paci, 2011), which assumes that the value of $w_{i j}$ is 1 if i and j share a border; otherwise, it is 0. The value of $w_{i j}$ should be standardized when calculating the value of $W ln {TF}_{i t}$ .

In equations (6) and (7), the $α_{z}$ s are the parameters that can be estimated by maximum likelihood estimation (MLE). A negative parameter value of $α_{z}$ indicates that the mean of production inefficiency of a region decreases as the value of $z$ increases, suggesting that the variable $z$ contributes to the improvement of regional production efficiency.

As a next step, we extend the specification of the production function and inefficiency equation by the time trend or time effects, as proposed by Bernini and Guizzardi (2010). The production function is augmented as follows

\begin{array}{l} ln y_{i t} = β_{0} + β_{k} ln K_{i t} + β_{l} ln L_{i t} + β_{k k} {(ln K_{i t})}^{2} + β_{k l} ln K_{i t} \times ln L_{i t} + β_{l l} {(ln L_{i t})}^{2} \\ \overset{}{} \overset{}{} \overset{}{} \overset{}{} \overset{}{} + β_{t} t + β_{t t} t \times t + β_{t l} t \times ln L_{i t} + β_{t k} t \times ln K_{i t} + (ν_{i t} - u_{i t}) \end{array}

Meanwhile, we add nine year-dummies as D2005–D2013 into the inefficiency function to capture the time trend. The duration of sampled data is from 2005 through 2014, and 2014 is set as a benchmark time in the inefficiency function.

The variance of the composite disturbance term is $δ^{2} = δ_{μ}^{2} + δ_{ν}^{2}$ , and let $γ = δ_{μ}^{2} / δ^{2}, γ \in [0, 1]$ . $γ = 0$ means that deviations from the optimal level of output can be completely explained by white noise; and $γ = 1$ suggests these deviations may be completely explained by the regional inefficiencies function. We assume that the joint distributions of $δ_{ν}^{2}$ and $δ_{μ}^{2}$ are normal–half-normal (Wang et al., 2006), and then use the MLE to estimate the econometric models.

Two different approaches apply to estimating models: the two-stage approach and the one-stage approach. In the former, we first estimate the inefficiency score of each unit, utilizing a stochastic frontier production function, and then establish a separate econometric model to estimate which factors determine the inefficiency score. In the second approach, the factors determining regional inefficiency are still included into an inefficiency function as in the two-stage approach. However, the inefficiency function is incorporated directly into the process of MLE and estimated simultaneously. Simar et al. (1994) propose the one-stage approach, which Battese and Coelli (1995) extended to the case of panel data. Wang and Schmidt (2002) demonstrate that the one-stage approach improves the estimate of the frontier parameters, and is superior to the two-stage approach. Therefore, this study uses the one-stage approach to estimate the model.

Data source and descriptive statistics

Table 2 shows detailed definitions of variables and database sources. Limited by data availability, the research period is restricted to 2005–2014. In several of those years, the data for patent application numbers for a few cities in Guizhou province were missing and we assume a stable growth rate for patent applications in these cities. We estimate the growth rate based on historical data, and then use the growth rate to make up the missing data. As a result, we obtain a panel data set with a total of 460 observations.

Table 2.

Definition of variables and data sources.

Variables	Definition	Unit	Source of the data
GDP _it	Region $i$ ‘s real GDPs at year t, based on the price level of 2004	Million Yuan	CEIC
K_it	The physical capital region $i$ accumulates at year t	Million Yuan	The data on the annual investment on physical capital are collected from CEIC
L_it	The number of labor (the number of people who are employed or running a business and whose age is not less than 16) region $i$ has at year t	Thousand persons	WIND financial database
TF _it	TF: tourist arrivals at region $i$ at year t, the sum of domestic and inbound tourism arrivals	Thousand person-times	CEIC
FDI _it	FDI region $i$ has at year t	Million dollars	CEIC
ATC _it	The accumulated technological capital, represented by the sum of the patent applications in region $i$ from the base year (2004) to year N	Number	The official websites of intellectual bureaus of regional governments
EE _it	The expenditure on education region $i$ at year t	Million Yuan	CEIC
PI _it	The proportion of industrial added value of region $i$ ‘s PI to region $i$ ‘s GDP at year t	Percent	The data on the industrial added value of the PI are collected from CEIC
PF _it	The proportion of region $i$ ‘s PF to region $i$ ‘s GDP at year t	Percent	The data on regional PF are collected from CEIC

Note: CEIC: China economy information database; FDI: foreign direct investment; PF: public finance; TF: tourist flow; GDP: gross domestic product; PI: primary industry. For a region, the stock of physical capital in a year is calculated by the perpetual inventory method. Accordingly, we adapted a calculation method proposed by Shan (2008) as Shan’s approach takes the specific context of China into account. Particularly, the stock of regional physical capital at year t = the investment of regional physical capital at year t + [the stock of regional physical capital at year t−1 $\times$ (1 − the discount rate of physical capital)]. The value of the discount rate is set as 10.96%. The stock of physical capital at the basic year (2004) = the investment of regional physical capital at year 2005/(the discount rate of physical capital + the average rate of physical capital investment growth). The values of GDP and K are adjusted based on the price level at 2004. According to CEIC, domestic tourism arrivals measure the number of Chinese (mainland) residents who travel to a particular city for sightseeing, vacation, visiting relatives, medical treatment, shopping, attending conference, or to engage in economic, cultural, sports, and religious activities; inbound tourism arrivals measure the number of foreigners, the citizens from Hong Kong, Macao, and Taiwan who come to a particular city in China (mainland) for sightseeing, vacation, visiting relatives, medical treatment, shopping, attending conference, or to engage in economic, cultural, sports, and religious activities.

Table 3 summarizes descriptive statistics. The big standard deviations of GDP, K and L, indicate a hugely unbalanced distribution of GDP, physical capital, and labor among the sample regions. The low levels of economic development in the sample regions are reflected by the low mean value and minimum of GDP, as well as by the high values of PI, which suggest that the regional economies of the sample regions depend heavily on their PIs. The maximum PF (1.54) comes from Aba, a Tibetan autonomous prefecture in Sichuan province, from the year 2010. Please note that in China’s administration system, an autonomous prefecture is an administrative region whose level is the same as that of a city. In some years, the PF values of Ganzi, another Tibetan autonomous prefecture in Sichuan province, were also higher than 1, which shows that in some years, regional GDP was too small to fund the operation of that region, or suggests that, without financial assistance from the central government or from other financial organizations, the local government could not function normally.

Table 3.

Summary of descriptive statistics.

Variable	Observation	Mean	Standard deviation	Minimum	Maximum
GDP	460	55396.54	77564.06	1889.71	752251.90
K	460	139209.60	229566.10	4437.99	2373251.00
L	460	206.83	125.25	3.20	820.60
TF	460	13638.80	16970.33	737.56	186208.00
FDI	460	1106.14	5646.69	0.06	69462.87
ATC	460	2672.51	14005.05	1.00	187405.00
EE	460	2454.09	2275.22	165.00	18305.52
PI	460	0.20	0.08	0.034	0.45
PF	460	0.29	0.20	0.08	1.54

Note: FDI: foreign direct investment; PF: public finance; TF: tourist flow; GDP: gross domestic product; PI: primary industry.

Results

The econometric models in this study are estimated using the FRONTIER 4.1 program. FRONTIER 4.1, developed by Coelli (1996), specializes in estimating stochastic frontier models and has been used extensively to deal with panel data in the literature, as Barros and Matias (2006), Hu et al. (2010), Bernini and Guizzardi (2010), and Kim (2011) demonstrate. FRONTIER 4.1 takes the MLE to obtain its results.

In Table 4, model (1) works as a benchmark for our analysis. For the production function, model (1) concludes with flexible output elasticity for the labor force and regional capital stock.

Table 4.

Estimation results.

	(1)	(2)	(3)
Stochastic frontier production function
$β_{0}$	11.812 (−8.09)***	11.115 (7.14)***	14.178 (7.25)***
$β_{L}$	1.522 (5.50)***	1.335 (4.72)***	2.201 (3.37)***
$β_{K}$	−1.272 (−4.40)***	−1.046 (−3.23) ***	−2.244 (−3.86)***
$β_{L K}$	−0.089 (−2.67)***	−0.075 (−2.23)**	−0.251 (−2.14)**
$β_{L L}$	−0.004 (−0.13)	0.002 (0.07)	0.083 (1.37)
$β_{K K}$	0.088 (5.03)***	0.074 (3.81)***	0.182 (3.33)***
			0.55 (5.46)***
			0.01 (4.93)***
			0.048 (1.76)*
			−0.071 (−3.36)***
Inefficiency function
$α$ ₀	3.839 (11.49)***	4.417 (11.35)***	5.854 (8.71)***
$α$ _lnTF	−0.067 (−2.59)***	−0.053 (−2.11)**	−0.055 (−1.84)*
$α$ _wlnTF		−0.101 (−3.06)***	−0.144 (−3.24)***
$α$ _lnFDI	−0.026 (−3.25)***	−0.032 (−3.93)***	−0.039 (−4.28)***
$α$ _lnATC	−0.103 (−7.05)***	−0.090 (−5.98)***	−0.093 (−5.31)***
$α$ _lnEE	−0.110 (−3.04)***	−0.078 (−2.14)**	−0.137 (−2.08)**
$α$ _lnPI	0.176 (4.70)***	0.158 (4.44)***	0.092 (1.52)
$α$ _lnPF	0.412 (11.04)***	0.425 (11.27)***	0.475 (8.98)***
			−1.132 (−7.53)***
			−0.971 (−7.32)***
			−0.811 (−6.98)***
			−0.78 (−7.22)***
			−0.693 (−7.15)***
			−0.567 (6.47)***
			−0.404 (−5.07)***
			−0.295 (−4.64)***
			−0.134 (−2.20)**
	0.057 (14.128)***	0.056 (14.340)***	0.057 (12.20)***
$γ$	0.636 (5.991)***	0.678 (5.906)***	0.9994 (227.10)***
Log likelihood function	25.402	29.693	49.326
LR test of the one-sided error	406.111	414.694	443.311

Note: FDI: foreign direct investment; PF: public finance; TF: tourist flow. t-Values are in parenthesis. All models are estimated by FRONTIER 4.1 software package.

*** Significance at the 1% level of statistics.

** Significance at the 5% level of statistics.

* Significance at the 10% level of statistics.

According to the inefficiency function in model (1), $lnTF$ has a significantly negative coefficient (−0.067), which demonstrates that the increase of regional tourist arrivals may result in a significant decrease in the gap between the frontier level and the real level of production. In model (1), the coefficients of other variables in the inefficiency function conform to our expectations: FDI and the accumulation technological capital (ATC) are positively related to regional production efficiency and the expenditure for education of a region (EE), reflecting that the level of regional human capital, contributes to improving production efficiency. We identify two negative determinants of regional production efficiency: the proportion of the PI’s contribution to regional GDP and the proportion of PF to regional GDP, and conclude that the presence of a primary regional industry and conservative behaviors by local government are associated with lower production efficiency.

The statistical significance of $δ^{2}$ suggests that the stochastic frontier model is appropriate. The high coefficient value of $γ$ (0.636) means the variability of the real production level can be explained mainly by the inefficiency function, which justifies the use of the stochastic frontier approach.

Model (2) includes the variable $W lnTF$ . The coefficient of $W lnTF$ (−0.101) in model (2) indicates a positive spatial effect of regional tourism on production efficiency, which means that tourism activities occurring in neighboring regions help to increase production efficiency of a particular region.

As an extension, we redesign the function specifications as model (3) to include time impacts or time trends. In model (3), the statistically significant coefficients of $β_{t}, β_{t t}, β_{t L}, β_{t K}$ indicate that time matters in the production functions; meanwhile, the statistically significant coefficients of $D_{2005} - D_{2013}$ indicate a clear year effect on variation of the average technical scores. The results of model (3) consistently demonstrate that tourism happening in the neighboring regions help lift regional productivity for a given region.

According to models (2) and (3), the absolute value of $α_{W lnTF}$ is greater than that of $α_{lnTF}$ . Therefore, we conclude that the spatial effect of tourism on regional productivity has a higher relevance than the local effect with regard to regional productivity. This conclusion seems to be counterintuitive as we often consider geographic distance a barrier inhibiting the flow of knowledge and information. However, the apparent inconsistency can be explained. In this study, a sample city usually has between two and six neighboring cities. Therefore, in most cases, the number of tourists in neighboring regions will be much greater than that in a local city itself. Table 5 shows a comparison of tourist arrivals in local cities and their neighboring cities for four sample cities, indicating a massive gap between local tourist arrivals and tourist arrivals for neighboring cities. Considering a larger base of tourists, we suggest that the knowledge (or information) nearby regions generate may be richer and more diverse than that derived in any single region. Neighboring regions may be more valuable resources for a city trying to lift its productivity. From this informed perspective, the geographic barrier seems not to be as strong as intuition suggests, as it does not inhibit knowledge and information flow across cities when they share a border.

Table 5.

Comparison between local tourist arrivals and tourist arrivals in neighboring cities: Data from four sample cities.

City		2005	2006	2007	2008	2009	2010	2011	2012	2013	2014
Ganzi	(1)	2308.20	2944.20	3265.00	1166.10	2700.20	3586.80	4376.90	5208.60	9585.50	8023.70
	(2)	11,990.03	17,936.40	23,564.90	21,545.80	28,156.60	36,281.50	47,721.80	59,842.40	107,564.20	70,784.48
Lincang	(1)	1331.73	1697.10	2271.90	2308.29	2535.01	2730.23	3043.20	3636.40	6679.60	5394.88
	(2)	9611.92	9889.79	14,003.00	11,957.37	14,479.20	16,964.88	21,052.40	26,707.90	47,760.30	53,794.30
Nujian	(1)	737.56	888.47	990.00	1209.01	1359.61	1505.44	1734.50	1913.30	3647.80	2446.53
	(2)	4805.961	4944.896	7001.5	5978.686	7239.602	8482.442	10,526.2	13,353.95	23,880.15	26,897.15
Ya’an	(1)	2963.23	4754.20	6504.40	7035.70	8002.90	10,054.20	11,064.00	11,952.70	23,016.70	10,731.60
	(2)	64,696.20	75,029.09	82,210.60	75,054.80	100,410.70	124,619.20	170,841.90	217,994.10	388,836.00	312,040.47

Note: TF: tourist flow. Rows (1) and (2) indicate local tourist arrivals (TF) and tourist arrivals in neighboring cities, respectively. The unit of tourist arrivals is one thousand person-times.

We calculate the efficiency scores of sample cities based on model (3), which controls well for both spatial effect and time trend. Table 6 reports the efficiency scores of sample cities in some years during the research period. It is clear that obvious heterogeneity in terms of efficiency scores among sample cities exists. On average, Kunming, Yuxi, South GZ, Chengdu, and Deyang reflect high levels of production efficiencies, whereas Nujiang, Tongren, and Ganzi show the lowest levels of production efficiencies. Table 6 indicates a decline trend for efficiency scores of many sample cities, which is in line with Wang and Tao (2010). The reason may lie in that during the research time period, many people with high-level skills or advanced knowledge in southeastern China chose to immigrate into the developed regions in China such as Zhejiang and Guangdong provinces, so that the average quality of human capital in southeastern China declines, and in turn the efficiency scores for the sample cities fall.

Table 6.

The efficiency scores of sample cities—based on the results of model (3).

Province	City	2005	2006	2009	2010	2013	2014	Province	City	2005	2006	2009	2010	2013	2014
Yunan	Kunming	0.983	0.948	0.836	0.784	0.762	0.715		Southeast GZ	0.239	0.236	0.225	0.258	0.253	0.245
	Qujing	0.531	0.532	0.496	0.475	0.443	0.393		South GZ	0.948	0.923	0.873	0.863	0.946	0.912
	Yuxi	0.906	0.895	0.853	0.828	0.736	0.675	Sichuan	Chengdu	0.995	0.985	0.958	0.975	0.966	0.944
	Baoshan	0.297	0.294	0.280	0.271	0.236	0.222		Zigong	0.725	0.735	0.669	0.662	0.580	0.524
	Shaotong	0.288	0.287	0.261	0.252	0.230	0.205		Panzhihua	0.780	0.777	0.753	0.740	0.658	0.626
	Lijiang	0.223	0.222	0.217	0.212	0.199	0.179		Luzhou	0.532	0.571	0.560	0.556	0.497	0.463
	Puer	0.266	0.257	0.239	0.232	0.207	0.191		Deyang	0.962	0.985	0.812	0.794	0.702	0.661
	Lincang	0.263	0.255	0.231	0.222	0.195	0.184		Mianyang	0.897	0.904	0.732	0.710	0.632	0.597
	Chuxiong	0.454	0.439	0.378	0.357	0.314	0.291		Guangyuan	0.372	0.367	0.305	0.294	0.269	0.253
	Honghe	0.480	0.471	0.401	0.372	0.337	0.311		Suining	0.581	0.572	0.515	0.486	0.433	0.401
	Wenshan	0.616	0.620	0.550	0.532	0.478	0.456		Neijiang	0.566	0.583	0.581	0.621	0.601	0.555
	Xishuangbanna	0.373	0.364	0.306	0.290	0.260	0.246		Leshan	0.590	0.584	0.630	0.615	0.573	0.518
	Dali	0.486	0.483	0.419	0.404	0.377	0.340		Nanchong	0.550	0.537	0.571	0.590	0.495	0.455
	Dehong	0.232	0.228	0.203	0.195	0.169	0.152		Meishan	0.517	0.508	0.499	0.494	0.462	0.428
	Nujiang	0.136	0.177	0.135	0.133	0.099	0.088		Yibin	0.621	0.616	0.627	0.624	0.553	0.500
	Diqing	0.210	0.215	0.192	0.193	0.182	0.208		Guangan	0.450	0.454	0.494	0.485	0.440	0.409
Guizhou	Guiyang	0.586	0.606	0.637	0.624	0.603	0.589		Dazhou	0.598	0.605	0.549	0.539	0.478	0.439
	Liupanshui	0.384	0.405	0.405	0.408	0.400	0.370		Yaan	0.431	0.421	0.418	0.405	0.360	0.342
	Zunyi	0.495	0.502	0.503	0.499	0.533	0.497		Bazhong	0.303	0.295	0.297	0.297	0.276	0.248
	Anshun	0.249	0.254	0.235	0.225	0.245	0.238		Ziyang	0.578	0.587	0.597	0.577	0.495	0.464
	Tongren	0.219	0.239	0.206	0.199	0.160	0.186		Aba	0.342	0.324	0.189	0.180	0.170	0.158
	Souwest GZ	0.362	0.363	0.339	0.326	0.362	0.350		Ganzi	0.200	0.204	0.177	0.172	0.150	0.132
	Bijie	0.310	0.312	0.302	0.294	0.434	0.355		Liangshan	0.519	0.510	0.491	0.480	0.429	0.392

Note: Due to the space limitation, the data of year 2007–2008 and 2011–2012 are dropped.

Conclusion and implications

To test the TLE proposition, this study employs panel data for 46 cities in southwestern China and then applies stochastic frontier production models with trans-log production specification to determine the relationship between tourism and regional production efficiency. This study concludes with the following empirical results:

For the sample cities in southwestern China, regional tourism, represented by the number of tourist arrivals, plays a positive role in lifting regional production efficiency. To ensure accurate estimations, we control for other determinants of regional production efficiency such as FDI, the accumulated technique capital, education investment, regional industrial structure, and the behavior paradigm of government. The impacts of other determinants are as expected.

There is a significant spatial effect of tourism on regional efficiency, which indicates that TFs in neighboring regions contribute to improvements in the production efficiency of a given region.

As for practical implications, this study suggests that regional organizations and governments in underdeveloped regions need to view regional tourism as an important chance to develop their economies. Underdeveloped regions should consider the enhancing production efficiency as a strategic value of tourism development since they usually lack alternative sources for adding new information or knowledge.

To take full advantage of regional tourism’s positive effects, local people and firms need to communicate and interact frequently and effectively with tourists visiting their regions, and regional leaders or business operators should proactively request tourists to provide feedback on services and products, and encourage tourists to invest in or migrate to the destination.

This study is also a call for interregional tourism cooperation, taking into account the positive spatial effect of tourism on regional production efficiency. In most cases, regions face limitations when building supply chains to ensure rich travel experiences. Tourism product cooperation among neighboring regions may guarantee a better experience for tourists and stimulate larger scale tourism. Consequently, tourism’s positive impacts on production efficiency may become more significant.

This study has certain limitations. Some variables affecting regional production efficiency are not included due to a lack of available data. These untested variables may result in endogenous issues in our empirical models. If data are available, it would be worthwhile to investigate whether the magnitude of tourism’s impact on regional productivity varies depending on the type of tourists. When data do become available, exploring tourism impacts on firms’ production efficiency can move forward. Such investigations at the firm level may make TLE proposition even more credible. Finally, although this study adopts the random effect technique to obtain results, future studies may consider the fixed effects variant and compare results.

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 work was supported by the Natural Science Foundation of China (NSFC) under project numbers 71773101 and 71172048.

References

Aigner

Lovell

Schmidt

(1977) Formulation and estimation of stochastic frontier production function models. Journal of Econometrics 6(1): 21–37.

Arbelo

Pilar Pérez-Gómez

Marta Arbelo-Pérez

(2017) Cost efficiency and its determinants in the hotel industry. Tourism Economics 23(5): 1056–1068.

Aschauer

(1989) Is public expenditure productive? Journal of Monetary Economics 23(2): 177–200.

Assaf

Alexander Josiassen

Haemoon Oh

(2016) Internationalization and hotel performance: The missing pieces. Tourism Economics 22(3): 572–592.

Avila

AFD

Evenson

(2010) Total factor productivity growth in agriculture: The role of technological capital. Handbook of Agricultural Economics 4: 3769–3822.

Baek

Lee

(2018) Searching for comparative value in small and medium-sized alternative accommodation: a synthesis approach. The Journal of Asian Finance, Economics and Business 5(2): 139–149.

Balaguer

Cantavella-Jorda

(2002) Tourism as a long-run economic growth factor: The Spanish case. Applied Economics 34(7): 877–884.

Barros

Matias

(2006) Assessing the efficiency of travel agencies with a stochastic cost frontier: A Portuguese case study. International Journal of Tourism Research 8(5): 367–379.

Battese

Coelli

(1995) A model for technical inefficiency effects in a stochastic frontier production function for panel data. Empirical Economics 20(2): 325–332.

10.

Bernini

Guizzardi

(2010) Internal and locational factors affecting hotel industry efficiency: Evidence from Italian business corporations. Tourism Economics 16(4): 883–913.

11.

Bonaglia

La Ferrara

Marcellino

(2000) Public capital and economic performance: Evidence from Italy. Giornale Degli Economisti E Annali Di Economia 60: 221–244.

12.

Brida

Risso

(2009) Tourism as a factor of long-run economic growth: An empirical analysis for Chile. European Journal of Tourism Research 2(2): 178.

13.

Chen

(2007) Applying the stochastic frontier approach to measure hotel managerial efficiency in Taiwan. Tourism Management 28(3): 696–702.

14.

Cheung

Ping

(2004) Spillover effects of FDI on innovation in China: Evidence from the provincial data. China Economic Review 15(1): 25–44.

15.

Coelli

(1996) A guide to FRONTIER version 4.1: A computer program for stochastic frontier production and cost function estimation. CEPA Working Paper 96/08. University of New England. Available at: http://www.uq.edu.au/economics/cepa/frontier.php (accessed 10 June 2018).

16.

Deprins

Simar

(1989) Estimating technical inefficiencies with correction for environmental conditions. Annals of Public and Cooperative Economics 60(1): 81–102.

17.

Durbarry

(2002) The economic contribution of tourism in Mauritius. Annals of Tourism Research 29(3): 862–865.

18.

Durbarry

(2004) Tourism and economic growth: The case of Mauritius. Tourism Economics 10(4): 389–401.

19.

Feng

Page

(2000) An exploratory study of the tourism, migration–immigration nexus: Travel experiences of Chinese residents in New Zealand. Current Issues in Tourism 3(3): 246–281.

20.

Fereidouni

Al-mulali

(2014) The interaction between tourism and FDI in real estate in OECD countries. Current Issues in Tourism 17(2): 105–113.

21.

Figini

Vici

(2010) Tourism and growth in a cross section of countries. Tourism Economics 16(4): 789–805.

22.

Filippini

Hrovatin

Zorić

(2008) Cost efficiency of Slovenian water distribution utilities: An application of stochastic frontier methods. Journal of Productivity Analysis 29(2): 169–182.

23.

Foxall

Johnston

(1987) Strategies of user-initiated product innovation. Technovation 6(2): 77–102.

24.

Helpman

Krugman

(1985) Market Structure and Foreign Trade. Cambridge: MIT Press.

25.

Chiu

Shieh

. (2010) A stochastic cost efficiency analysis of international tourist hotels in Taiwan. International Journal of Hospitality Management 29(1): 99–107.

26.

Joshi

Sharma

(2004) Customer knowledge development: Antecedents and impact on new product performance. Journal of Marketing 68(4): 47–59.

27.

Kim

(2011) Factor determinants of total factor productivity growth in the Malaysian hotel industry: A stochastic frontier approach. Cornell Hospitality Quarterly 52(1): 35–47.

28.

Kim

Lim

Park

(2007) The effect of imports and exports on total factor productivity in Korea. Research Institute of Economy, Trade and Industry Discussion Paper Series, (07-E).

29.

Kireyeva

Abilkayir

Tsoy

(2018) A study on the distribution of information and high technology clusters: Kazakhstan’s experience. The Journal of Distribution Science 16(4): 1–12.

30.

Lee

Syah

(2018) Economic and environmental impacts of mass tourism on regional tourism destinations in Indonesia. The Journal of Asian Finance, Economics and Business 5(3): 31–41.

31.

Marrocu

Paci

(2011) They arrive with new information. Tourism flows and production efficiency in the European regions. Tourism Management 32(4): 750–758.

32.

Navaretti

Tarr

(2000) International knowledge flows and economic performance: A review of the evidence. The World Bank Economic Review 14(1): 1–15.

33.

Oliveira

Pedro

Marques

(2014) Cost efficiency of Portuguese hotels in the Algarve: A comparative analysis using mathematical and econometric approaches. Tourism Economics 20(4): 797–812.

34.

Shan

(2008) Reestimating the capital stock of China: 1952–2006. Journal of Quantitative & Technical Economics 10(1): 17–31.

35.

Simar

Lovell

van den Eeckaut

(1994) Stochastic frontiers incorporating exogenous influences on efficiency. Discussion Papers 9403.

36.

Sinclair

(1998) Tourism and economic development: A survey. The Journal of Development Studies 34(5): 1–51.

37.

Singh

(2008) Small island developing states (SIDS): Tourism and economic development. Tourism Analysis 13(5–6): 629–636.

38.

Temple

(2001) Generalizations that aren’t? Evidence on education and growth. European Economic Review 45(4): 905–918.

39.

Von Hippel

(1978) A customer-active paradigm for industrial product idea generation. Research Policy 7(3): 240–266.

40.

Wang

Gong

Cheng

(2006) China’s regional differences in technical efficiency and the decomposition of total factor productivity growth (1978–2003). Social Sciences in China 2: 55–66.

41.

Wang

Schmidt

(2002) One-step and two-step estimation of the effects of exogenous variables on technical efficiency levels. Journal of Productivity Analysis 18(2): 129–144.

42.

Wang

Tao

(2010) An empirical study on regional productivity and its determinants. System Engineering - Theory & Practice 30(10): 1762–1773.

43.

Lee

(2017) Use intention of mobile fingerprint payment between UTAUT and DOI in China. The Journal of Distribution Science 15(10): 15–28.

44.

(2000) Is China’s economic growth sustainable? A productivity analysis. China Economic Review 11(3): 278–296.

45.

Young

(1991) Learning-by-doing and the dynamic effects of international trade. Quarterly Journal of Economics 106(2): 369–405.