Specialization versus diversity: The role of city size in tourist cities

Abstract

The paper examines the impact of specialization and diversity on economic growth of tourist cities. Industrial diversity is decomposed into two parts, related and unrelated diversity (UV). The paper constructs a panel threshold model with city size as the threshold variable, and empirically tests the nonlinear effects of specialization and diversity on economic growth of tourist cities from the perspective of city size. The results show that: First, tourism specialization has a positive impact on economic growth of tourist cities, and there is a double threshold effect based on city size. When the city size exceeds two thresholds, tourism specialization plays a significant role in promoting economic growth of tourist cities. Second, related diversity (RV) has a positive impact on economic growth of tourist cities, and there is no threshold effect based on city size. Third, the impact of unrelated diversity on economic growth of tourist cities has a single threshold effect based on city size. Specifically, when the city size is lower than the threshold value, unrelated diversity inhibits economic growth of tourist cities. Otherwise, unrelated diversity will promote economic growth of tourist cities. The conclusions are still valid after a series of robustness tests. The paper highlights that local governments of tourist cities should not pursue absolute tourism specialization or industrial diversity, but should design and adjust industrial structure reasonably according to the city size.

Keywords

tourism specialization related diversity unrelated diversity tourist cities city size

Introduction

The existence of cities benefits from increasing returns to scale of agglomeration economies (Krugman, 1991). According to new growth theory and new economic geography theory, Marshall-Arrow-Romer (MAR) externalities (specialization) and Jacobs externalities (diversity) are two main sources of industrial agglomeration effect. The former is an externality effect occurring within industries, emphasizing that specialized agglomeration of similar industries is conducive to knowledge spillover and urban growth (Arthur, 1920; Arrow, 1962; Romer, 1986, 1990). The latter is an external effect occurring between industries, focusing on the coexistence of multiple industries can bring vitality to the city more than a single industry (Feldman and Audretsch, 1999; Glaeser et al., 1992; Jacobs, 1969, 1984).

Previous studies have focused on the important role of tourism specialization on economic growth (i.e., MAR externalities) (Chang et al., 2012; Croes 2013; Croes et al., 2018, 2021; De Vita and Kyaw, 2017; Jackman, 2014; Lanza et al., 2003; Marsiglio, 2018; Soulie and Valle, 2014). Despite its importance, tourism is one of the most vulnerable sectors (Dogru et al., 2019; Duro et al., 2021; Liu and Pratt, 2017). “Black swan” incidents, such as the financial crisis in 1997 and 2008, the severe acute respiratory syndrome (SARS) epidemic in 2003, and various earthquake and social unrest events, have brought significant negative impacts on tourism (Zhang et al., 2021). Especially since the end of 2019, the global tourism industry has been hit hard again by the COVID-19 pandemic (Gil-Alana and Poza, 2020; Gssling et al., 2021; Kreiner and Ram, 2020). As a result, the economies of countless tourist cities have fallen into recession (Lim and To, 2021). It can be seen that the urban economy relying solely on tourism specialization is unstable. In the context of the pandemic, how to promote stable economic growth of tourist cities has become a crucial research issue.

In recent years, a few scholars have begun to pay attention to the effect of industrial diversity on the development of tourist cities (i.e., Jacobs externalities). For example, Sheng (2011) pointed out that tourist cities should actively use the revenue from booming tourism to pursue economic diversity, so as to avoid problems such as Dutch disease caused by the rapid growth of tourism. Based on the research of Frenken et al. (2007), Aarstad et al. (2016) divided industrial diversity into two dimensions of related diversity (RV) and unrelated diversity (UV), and believed that rich RV and limited UV were most conducive to the development of tourism destinations. Erkuş-Ozturka & Terhorst (2018) showed that the development of tourism stimulated the growth of related and unrelated industries and generated economic diversity. On the whole, these studies provide a new perspective for economic growth of tourist cities. However, the role of diversity in economic growth of tourist cities still lacks empirical tests at the regional level.

As the two main sources of industrial agglomeration, the externalities of specialization and diversity are significantly different (Attaran, 1986; Batisse, 2002; Glaeser et al., 1992; Henderson et al., 1995). Therefore, in order to ensure stable economic growth, should tourist cities focus on developing tourism specialization or promote the diversified development of industries? According to the literature on the relationship between tourism and economic growth, the research conclusions will be different depending on the sample range selected by scholars. Most scholars support the tourism-led growth hypothesis (TLGH); they believe that tourism can drive economic growth (Brau et al., 2007; Chang et al., 2012; Croes 2013; Lanza et al., 2003). However, a small number of scholars have found that tourism specialization has no significant impact on economic growth (Croes et al., 2021) and may even have a certain negative impact in economically unstable regions (Jackman, 2014). Therefore, it is difficult for us to directly judge which externalities economic growth of tourist cities comes from. Especially for China, which has a large number of cities, the size of tourist cities varies greatly, from small cities with a population of less than 500,000 to megacities with a population of more than two million. It remains to be studied which industrial agglomeration mode (specialization or diversity) should be chosen by tourist cities of different sizes.

Using the panel data of 60 typical tourist cities in China from 2005 to 2019, the paper adopts a nonlinear threshold regression model to empirically test the impact of specialization and diversity on economic growth of tourist cities from the perspective of city size. The main contributions of this paper are as follows:

First, existing studies have mainly focused on the impact of tourism specialization on the city’s economic growth, and less attention has been paid to the important role played by industrial diversity. Therefore, this paper uses entropy method to decompose industrial diversity into two parts, related and unrelated, and discusses their impact on economic growth of tourist cities. Second, previous literature on the relationship between tourism and economy neglected the impact of city size. The paper constructs a panel threshold model with city size as the threshold variable and empirically tests the nonlinear effects of specialization and diversity on economic growth of tourist cities from the perspective of city size.

The remainder of the paper is structured as follows: Literature review and research hypothesis introduces the literature review and research hypothesis; Research design explains the research design, including empirical models, variables, and data; Empirical results and analysis presents the empirical results and analysis; and Conclusions and implications is the conclusions and implications.

Literature review and research hypothesis

Theoretical background

Specialization externality and diversification externality

Classical economists represented by Romer regard technological externalities as the engine of economic growth (Romer, 1986). Enterprise agglomeration facilitates the dissemination of knowledge and information, especially tacit knowledge, promotes knowledge spillover among enterprises, and boosts economic growth. According to the theory of agglomeration economy, agglomeration externalities are divided into specialization externalities and diversification externalities. Specialization externalities refer to those generated by enterprises in the same industry. Marshall (1920) believed that enterprises in the agglomeration area acquired externalities due to sharing “reservoir” of labor force, intermediate input of non-tradable goods, and knowledge spillover, thus forming industrial agglomeration. Industrial agglomeration further promotes division of labor and is conducive to specialization of production. Subsequently, Arthur (1920), Arrow (1962), and Romer (1986, 1990) studied the impact of externalities on economic growth and argued that production specialization is beneficial to technical progress and production efficiency. Therefore, based on the viewpoints of the above scholars, the agglomeration externalities formed by the enterprises in the same industry are called Mar-Arrow-Romer externalities (MAR externalities), also known as specialization externalities.

The viewpoint of diversified externalities comes from the famous urban economist Jacobs (1969, 1984). He emphasized that knowledge spillovers originated from enterprises in different industries, that is, diversification promotes economic growth. This view is called Jacobs externalities, also known as diversification externalities. In a specific region, industrial diversification is conducive to the activities of inter-industry enterprises and promotes the exchange of knowledge, technology, information, and technological innovation, thus boosting the development of regional economy (Dissart, 2003; Skyes, 1950). In addition, urban diversification also enables enterprises to have more stable investment choices. When the price of a certain upstream product fluctuates, they can choose alternative products of other industries as new sources of investment (Neffke et al., 2011), thus promoting urban development.

Related diversity and unrelated diversity

With the expansion of theoretical research, scholars have further deepened their understanding of industrial diversification. Early literature believed that the level of industrial diversification was measured by the degree of equilibrium of industrial distribution and the number of types of industries. This view ignores the interconnectedness of different industries. Siegel et al. (1995) pointed out that, in addition to regional economic scale, industrial diversification is also related to the interconnection between industries. Wagner and Deller (1998) believed that the traditional definition of industrial diversification emphasizes the distribution of employment across industries but ignores elements that fully reflect the linkages between industries. Frenken et al. (2007) firstly decomposed industrial diversification into RV and UV. Specifically, RV represents the level of diversification among industrial sectors with complementary abilities and moderate cognitive distance, and UV represents the level of diversification among industries without technological linkages. Knowledge spillover is the main mechanism of RV. The relevant and diversified environment is conducive to the re-integration of scattered and diversified knowledge into new knowledge (Boschma, 2005). The industrial sectors involved in UV do not share complementary capabilities, and lack input-output linkages and opportunities for knowledge sharing, so it is difficult to produce effective knowledge spillover. But industry irrelevance helps spread the risk from changes in external demand. If one industry sector is hit by an economic crisis, cities can spread the risk by developing other unrelated industries. Since there is no input-output connection between industries, the impact of the crisis will not be transmitted to other unrelated industries, thus maintaining employment and economic stability (Frenken et al., 2007).

The impact of tourism specialization on economic growth

Existing literature on the relationship between tourism specialization and economic growth is still controversial. Most scholars support the TLGH and agree that tourism specialization can drive economic growth. At the national level, Lanza et al. (2003) took 13 OECD countries as examples and found that tourism specialization would not harm economic welfare. Brau et al. (2007) pointed out that if small-scale countries actively pay attention to the development of tourism, it will greatly promote economic growth. Based on the data from 159 countries, Chang et al. (2012) found a positive relationship between economic growth and tourism. Croes (2013) showed that tourism specialization was not harmful to economic growth of small islands and could make up for technological gaps and resource constraints. At regional level, Jin (2011) demonstrated the positive short-run effect of tourism on Hong Kong’s economic growth. Lee (2012) found that tourism has indirect effects on economic growth in the long run in Singapore. Brida and Giuliani (2013) indicated that the TLGH holds for South Tyrol and Trentino (sub-national trans-frontier economies). Wang and Xia (2013) concluded that there is a long-term integration relationship between economic growth of Gaochun and tourism revenue. Neuts (2020) confirmed a long-term equilibrium between tourism and gross domestic product (GDP) in German cities. However, a few scholars also point out that the development of tourism cannot bring sustainable economic growth. For example, Jackman (2014) believed that tourism specialization may have negative effects in areas where tourism is particularly unstable. Croes et al. (2018) and Croes et al. (2021) found that tourism specialization had no significant direct impact on economic growth.

Although these studies help us to understand the relationship between tourism and economic growth, it is worth considering that the impact of tourism on economic growth may be different under different levels of conditional factors. In other words, it is possible that the current conclusions cannot be agreed because there are nonlinear effects between them. A few studies have explored the nonlinear effect of tourism specialization and economic level. For example, Sequeira and Nunes (2008) found that poor countries always benefit from tourism specialization. Chang et al. (2012) pointed out that tourism specialization has a greater impact on economic growth in tourist destinations with a lower level of economic development. Deng et al. (2014) confirmed a non-monotonic relationship between international tourism and economic growth. They further pointed out that tourism-led economic growth may not be sustained at high levels of tourism specialization. De Vita and Kyaw (2017) believed that the positive impact of tourism specialization on economic growth depends on the level of economic development and the absorption capacity of the financial system.

Recently, some studies have concluded that city size appears to be an important determinant of tourism specialization for economic growth. At the national level, the positive correlation between tourism specialization and economic growth seems to be more obvious in small-scale countries (Brau et al., 2007; Croes, 2013; Jackman, 2014). At the city level, the larger the city size is, the more tourism development and economic benefit will be improved (Huang et al., 2000; Zhang and Sun, 2016). In general, in tourist cities with smaller size, tourism specialization may result in a monolithic economy and over-reliance on tourism. Furthermore, the influx of tourists may put pressure on cities with limited capacity and cause serious social problems. However, large-scale tourist cities often have complete infrastructure systems, systematic tourism planning and development capabilities, and stable consumer markets. These advantages can enhance the attractiveness of cities to tourists and promote the aggregation of advantageous resources and factor allocation (Wu et al., 2020). Therefore, tourism specialization seems to be desirable in large-scale tourist cities. Based on this, the paper proposes the first research hypothesis:

H1: The impact of tourism specialization on economic growth of tourist cities is closely related to city size.

The impact of industrial diversity on economic growth

At present, there is no consensus on the effects of diversity on economic growth. Most scholars support the idea that diversity promotes economic growth. For example, Attaran (1986) believed that diversity is conducive to reducing the impact of fluctuations caused by changes in economic factors outside the region on economic growth. Glaeser et al. (1992) found that industrial diversity had a significant impact on economic growth, while industrial specialization had no significant impact. Batisse (2002) showed that industrial diversity promoted economic growth, while industrial specialization inhibited economic growth. Simonen et al. (2015) proposed that the diversified structure of high-tech sector did play a role in regional economic growth. However, a few scholars have found that diversity has no significant impact on economic growth. For example, Henderson (1986) pointed out that industrial diversity had no significant impact on small and medium-sized regional economic clusters. Fan et al. (2014) showed that diversified economy had no significant impact on total factor productivity. Although there is still a lack of empirical tests on the relationship between diversity and economic growth of tourist cities, existing studies all support the view that diversity is beneficial to economic growth of tourist cities (Aarstad et al., 2016; Erkuş - Ozturka & Terhorst, 2018; Sheng, 2011).

It is worth noting that some studies have found that city size may influence economic growth effect of diversity. Duranton et al. (2001) discussed the mechanism for manufacturing enterprises to relocate from metropolitan to small and medium-sized cities. The results show that enterprises often take advantage of the diversified environment of big cities to develop new production processes, and when the production technology is gradually mature and standardized, enterprises will relocate to specialized small- and medium-sized cities. Duranton and Puga (2005) further pointed out that cities of different sizes will move toward functional division of labor, such as large- and medium-sized cities represented by headquarters and advanced business services, small- and medium-sized cities dominated by ordinary manufacturing and processing industries. Sun and Zhou (2013) found that specialized agglomeration in small cities has a significant economic growth effect, while diversified industrial agglomeration in big cities has a significant economic growth effect. Su and Zhao (2011) showed that the larger the city size, the higher the level of industrial diversity. The expansion of city size will produce agglomeration economic benefits. When the city’s development reaches a certain threshold scale, it may be able to benefit from the linkage of upstream and downstream industries, and promote productivity improvement and economic growth (Ke and Zhao, 2014). Therefore, the impact of industrial diversity on regional economy may be nonlinear due to the city size. Based on the above analysis, we propose the second research hypothesis:

H2: The impact of diversity on economic growth of tourist cities is closely related to city size.

Existing empirical evidence shows that the impact of RV and UV on economic growth is different. Frenken et al. (2007) studied the Netherlands from 1996 to 2002 and found that RV had a significant positive impact on regional employment and productivity growth, while UV had no significant impact. Based on a study of 205 European regions, Van Oort et al. (2015) found that RV was significantly correlated with employment growth, especially in small and medium-sized urban areas, while UV was not. Fritsch and Kublina (2018) pointed out that the West German region benefited from the effects of both related and UV, but the effects of UV were more pronounced. In tourism research, Aarstad et al. (2016) first paid attention to the impact of related and UV on tourist destinations and believed that rich RV and limited UV were the best choices for cities. However, they did not provide empirical evidence.

A tourist city can be regarded as a production system. From the perspective of tourists, the services or products in this system are usually provided by many suppliers (Haugland et al., 2011). RV means that there are many suppliers offering different but complementary products. On the one hand, enterprises in related industries also share resources, information, and knowledge to reduce costs, thus improving the production efficiency of enterprises and industries. On the other hand, RV can promote knowledge spillover between complementary industries. The exchange of knowledge between relevant departments can promote innovative search behavior and diffusion of new technologies, so that tourist cities can provide novel products or services and promote regional economic growth.

Unrelated diversity is an industrial pattern in which industries without obvious technical and economic ties are distributed in a particular region. Its impact mechanism on city growth is different from RV. On the one hand, co-production may be limited in an undiversified region due to the lack of local complementarity of products and services. In addition, unrelated industries will fragment the knowledge base, which may further limit the development of integrated tourism products, thus detrimental to the growth of the city’s economy. On the other hand, from the perspective of externalities, unrelated industrial agglomeration can improve the matching opportunities of material capital, human capital, technology, and other elements required by enterprises in the process of production and operation. This can not only effectively reduce the search cost of factor supply and demand but can also help to increase the types of intermediate inputs, thus improving the production efficiency of enterprises and promoting regional economic growth. Therefore, we propose the third research hypothesis:

H3: The impact of RV on economic growth of tourist cities is different from that of UV on economic growth of tourist cities.

Research design

Empirical models

In order to test the hypothesis of this paper, we use the threshold regression model and take city size as the threshold variable to test the nonlinear relationship between tourism specialization, industrial diversity (related and unrelated), and economic growth of tourist cities, respectively.

Before threshold regression, whether the sample has threshold effect needs to be tested. We first conduct a single threshold test on the model. If the result is not significant, it indicates that there is no threshold effect, and we use a linear model to estimate. On the contrary, if the single threshold test results are significant, the double threshold and triple threshold tests are continued until the results are not significant. Then, in the case of threshold effect, the threshold value level of each threshold is further estimated, and the parameter estimation result of threshold regression is finally reported. Table 1 shows the significance test, threshold estimation, and confidence interval of city size thresholds.

Table 1.

Threshold effect of city size.

Key dependent variables	Threshold type	F stat	Critical value			Threshold value	Confidence interval
Key dependent variables	Threshold type	F stat	10%	5%	1%	Threshold value	Confidence interval
Tourism specialization	Single	58.96^*	51.43	58.94	69.12	6.6481^*	(6.6384, 6.6542)
	Double	51.11^**	43.31	49.38	66.80	6.6481^**	(6.6384, 6.6542)
						7.6350^**	(7.5959, 7.6689)
	Triple	27.85	35.82	48.26	140.85
Related diversity	Single	12.25	21.70	28.74	43.94
Unrelated diversity	Single	66.41^**	40.14	48.33	60.19	6.6481^**	(6.6403, 6.6542)
	Double	16.47	31.89	41.40	63.47
	Triple	22.21	30.76	38.64	79.69

Notes: The F value and the critical values of 10%, 5%, and 1% are obtained by repeated sampling 300 times using “self-sampling.” ***, **, and * denote significant at the 1%, 5%, and 10% levels, respectively.

According to Table 1, first, the single threshold and double threshold of tourism specialization are significant at the 5% level, and the triple threshold effect does not pass the significance test. This shows that when city size is taken as the threshold variable, the model with tourism specialization as the key independent variable rejects the assumption of linear relationship, and tourism specialization has a double threshold effect on economic growth of tourist cities. Therefore, we construct a double threshold model between tourism specialization and economic growth of tourist cities as follows

\ln {P G D P}_{i t} = β_{0} + β_{1} {l n T R}_{i t} * I ({l n P O P}_{i t} \leq γ_{0}) + β_{2} {l n T R}_{i t} * I (γ_{0} < {l n P O P}_{i t} \leq γ_{1}) + β_{3} {l n T R}_{i t} * I ({l n P O P}_{i t} > γ_{1}) + β_{4} X_{i t} + μ_{i t}

(1)

where i indexes city, t denotes year, and

{l n P G D P}_{i t}

represents economic growth level of tourist city;

{l n T R}_{i t}

is the key independent variable, representing the level of tourism specialization; and

ε_{i t}

denotes the random error term. City size (lnPOP) is the threshold variable, γ is the threshold value to be estimated, and

Ι (\cdot)

indexes the indicator function. If the requirements in brackets are met, the value is 1; otherwise, it is 0.

X_{i t}

represents a series of control variables, including human capital (

{l n E D U}_{i t}

), infrastructure (

{l n I N F}_{i t}

), industrial structure (

{l n I N D}_{i t}

), government intervention (

{l n G O V}_{i t}

), and openness (

{l n F D I}_{i t}

). All variables are transformed into natural logarithm to reduce data variability and the possible influence of heteroscedasticity.

Second, the single threshold effect of RV does not pass the significance test, indicating that there is no threshold effect of RV on economic growth of tourist cities Therefore, we construct the following empirical model to investigate the impact of RV on economic growth of tourist cities

{l n P G D P}_{i t} = β_{0} + β_{1} {l n R V}_{i t} + β_{2} X_{i t} + ε_{i t}

(2)

where

{l n R V}_{i t}

is the key independent variable, indicating the level of RV. The remaining variables have the same meaning as model (1).

Third, the single threshold of UV is significant at the 5% level, and neither the double threshold nor triple threshold effects pass the 10% significance test. The results show that the UV has a single threshold effect on economic growth of tourist cities when the city size is taken as the threshold variable. Therefore, we construct a single threshold model between UV and economic growth of tourist cities as follows

\ln {P G D P}_{i t} = β_{0} + β_{1} {l n U V}_{i t} * I ({l n P O P}_{i t} \leq γ_{2}) + β_{2} {l n U V}_{i t} * I ({l n P O P}_{i t} > γ_{2}) + β_{3} X_{i t} + μ_{i t}

(3)

where

{l n U V}_{i t}

is also the key independent variable, indicating the level of UV. The remaining variables have the same meaning as model (1).

Variables

The dependent variable

Existing literatures mostly adopt GDP or per capita GDP as the variable of city’s economic growth (Frenken et al., 2007; Sun and Chai, 2012). Considering the difference of city’s population size, this paper uses the per capita GDP of cities over the years as the dependent variable and converts it into constant price based on 2005.

The key independent variables

Tourism specialization

Referring to the studies of Adamou and Clerides (2010), Brau et al. (2007), Deng et al. (2014), and Zhang et al. (2020), in the paper, tourism specialization level (TR) is measured by the proportion of the nominal tourism revenue (i.e., the sum of domestic tourism revenue and inbound tourism revenue) in the nominal GDP. Among them, the inbound tourism revenue is converted into RMB using the average exchange rate of RMB against US dollar in the current year.

Diversity index decomposition based on entropy index: related and unrelated diversity

Drawing on the practice of Frenken et al. (2007), the paper uses the measurement method of entropy index to decompose industrial diversity (VAR) into RV and UV. The overall diversity of the industry can be expressed as

E_{T} = \sum_{j = 1}^{n} P_{j} \ln (1 / P_{j})

(4)

where

E_{T}

is the entropy value, representing the level of industrial diversity.

P_{j}

is the proportion of employment in industry j to the total employment in the local city. If a city has only one industrial sector, the entropy value is 0. Conversely, the higher the entropy value is, the higher the degree of industrial diversity is.

In order to illustrate the decomposition process of related and UV, we assume that there are S major sectors in the economic system, and S major sectors can be subdivided into n subsectors (n ≥ S). For one of the major sectors, that is s, its proportion of employment is equal to the sum of the proportion of employment in each small sector, which can be expressed as

P_{s} = \sum_{j \in s} P_{j}

(5)

Then, the degree of diversity of each subdivision within a major sector s can be expressed as follows

E_{w} = \sum_{j \in s} (P_{j} / P_{s}) \ln (P_{s} / P_{j})

(6)

The diversity entropy index among each major sector s is

E_{A} = \sum_{s = 1}^{S} P_{s} \ln (1 / P_{s})

(7)

Based on the measurement method of entropy index, the overall industrial diversity is decomposed as follows

E_{T} = \sum_{s = 1}^{S} \sum_{j \in s} P_{j} \ln (1 / P_{j}) = \sum_{s = 1}^{S} [\sum_{j \in s} (P_{s} / P_{s}) P_{j} (\ln (P_{s} / P_{j}) + \ln (1 / P_{s}))] = \sum_{s = 1}^{S} P_{s} (\sum_{j \in s} (P_{j} / P_{s}) \ln (P_{s} / P_{j}) + (\sum_{s = 1}^{S} P_{s} \ln (1 / P_{s})) = \sum_{s = 1}^{S} P_{s} E_{w} + E_{A}

(8)

where

\sum_{s = 1}^{S} P_{s} E_{w}

is the product of the degree of internal diversity (

E_{w}

) and the proportion of employment (

P_{s}

) in a major sector s. It measures the degree of diversity of various subsectors within a major sector s, known as RV.

E_{A}

refers to the degree of diversity among each major sector s in an economic system, which represents the level of sectoral diversity with relatively weak industrial correlation, that is, UV.

Industrial classification method

When the entropy index method is used to calculate related and UV indexes, it is necessary to clarify the standards of industry classification at different levels. As for the division of small sectors, the paper follows the Industrial Classification for National Economic Activities (GB/T 4754-2002) and divides the industry into 19 small industrial sectors. Specifically, it includes: agriculture, forestry, animal husbandry, and fishery; mining; manufacturing; electricity; gas and water production and supply; construction; transportation, storage, and postal services; information transmission, computer services, and software; wholesale and retail; accommodation and catering; finance; real estate; leasing and commercial services; scientific research, technical services, and geological survey; water conservancy, environment, and public facilities management; residential and other services; education, health, social security, and social welfare; culture, sports and entertainment; public administration; and social organizations.

The division of major sectors is mainly based on the three industries. Among them, the primary industry is agriculture, forestry, animal husbandry, and fishery. The secondary industry includes mining, manufacturing, electricity, gas and water production and supply, and construction. As there are many subsectors in the tertiary industry, further classification is needed. This paper refers to Browning and Singlemann (1975) and classifies the service industry into four categories according to the United Nations Standard Industrial Classification (SIC): producer service, consumer service, circulation service, and social service. To be specific, producer services include finance, real estate, leasing, and business services. Consumer services include accommodation and catering, residential and other services, culture, sports, and entertainment. Circulation services include transportation, warehousing and postal services, information transmission, computer services and software, and wholesale and retail. Social services include scientific research and technical services and qualification exploration, water conservancy environment and public facilities management, education, health and social security and social welfare, public administration, and social organizations.

Control variables

In addition to tourism specialization and industrial diversity, there are some other factors affecting economic growth of tourist cities. Therefore, referring to existing literature, this paper takes human capital, infrastructure, industrial structure, government intervention, and openness as control variables. Details are as follows:

(1) Human capital (EDU): the number of college students per ten thousand is used as the proxy variable of human capital.

(2) Infrastructure (INF): it is expressed by the per capita road area of the city.

(3) Industrial structure (IND): it is measured by the proportion of the GDP of the secondary industry in a city.

(4) Government Intervention (GOV): it is measured by the proportion of the city’s government fiscal expenditure to local GDP.

(5) Openness (FDI): it is expressed by the proportion of foreign investment actually utilized in the city to local GDP. For the foreign investment denominated in US dollars, the average exchange rate of the year is converted into RMB.

Threshold variable

City size is an important factor affecting the externality of industrial agglomeration. From the perspective of economic and social development, the aggregation of population is generally corresponding to the comprehensive development level of the city (Li and Wang, 2014). Therefore, this paper measures the city size (POP) by the number of permanent residents in the city.

Data

The paper conducts an empirical analysis on the balanced panel data of 60 typical tourist cities in China from 2005 to 2019. The data comes from China City Statistical Yearbook, China Tourism City Yearbook, and National Economic and Social Development Statistical Bulletin. In order to get balanced panel data, the mean value is used to fill in a few missings. The descriptive statistics of all variables are shown in Table 2. To avoid the bias caused by outliers and ensure the accuracy of the results, all continuous variables are winsorized at the 1% and 99% levels as robustness test.

Table 2.

Descriptive statistics.

	Variable	Symbol	N	Mean	SD.	Min	Max
Dependent variable	Economic growth	lnPGDP	900	10.59	0.66	8.52	12.29
Key dependent variable	Tourism specialization	lnTR	900	2.74	0.72	0.34	5.64
	Related diversity	lnRV	900	−0.01	0.18	−1.88	0.23
	Unrelated diversity	lnUV	900	0.26	0.12	−0.82	0.56
Control variables	Human capital	lnEDU	900	5.68	0.94	2.20	7.70
	Infrastructure	lnlNF	900	1.61	0.73	−0.69	5.40
	Industrial structure	lnIND	900	3.76	0.26	2.69	4.40
	Government intervention	lnGOV	900	2.61	0.46	0.02	4.08
	Openness	lnFDI	900	0.58	1.29	−6.44	2.66
Threshold variable	City size	lnPOP	900	6.22	0.80	3.81	8.05

Empirical results and analysis

Panel unit root test and cointegration test

In order to ensure the validity of the estimation results and avoid the problem of pseudo regression, this paper uses panel unit root test to examine whether the panel data is stationary. The methods of panel unit root test can be divided into two types according to whether there is cross-sectional correlation in the panel data. One is the first-generation conventional panel unit root tests based on the hypothesis of “cross-sectional independence,” such as LLC test, HT test, IPS test, ADF test, and PP test. The other is the second-generation panel unit root test (the CIPS test) proposed by Pesaran (2007) based on the hypothesis of “cross-section dependence.” If there is cross-sectional dependence in the panel data but the first-generation panel unit root tests are adopted, it is likely to reject the null hypothesis excessively. Therefore, the paper first looks at whether cross-sectional dependence exists within the panel data by applying Pesaran’s CD test (Pesaran, 2004) before the panel unit root analysis.

The results of the CD test are reported in Table 3. In column (1), we have strong evidence to reject the null hypothesis of cross-sectional independence for all of the variables. In other word, there is significant cross-sectional dependence in all variables. Therefore, the paper applies the CIPS test as the unit root test method, and the results are shown in Table 3. In column (2), lnEDU, lnINF, and lnIND reject the null hypothesis of the existence of unit root at the significance level of 5%, while other variables cannot reject the null hypothesis and are non-stationary. Therefore, we do the unit root test for the first order difference of all variables. As can been seen from column (3), all the variables in the first-order difference sequence reject the null hypothesis of the existence of unit root at the significance level of 1%, indicating that the panel data in this paper is integrated of order 1.

Table 3.

Results of cross-sectional dependence test and panel unit root test.

Variable	CD test	CIPS test
Variable	CD test	Level	∆
lnGDP	162.927^***	−2.060	−2.974^***
lnTR	161.070^***	−1.753	−2.747^***
lnRV	8.764^***	−1.985	−3.475^***
lnUV	140.935^***	−1.885	−3.395^***
lnEDU	162.537^***	−2.146^**	−3.324^***
lnINF	143.039^***	−2.343^***	−3.831^***
lnIND	162.505^***	−4.217^***	−4.620^***
lnGOV	162.078^***	−1.587	−3.325^***
lnFDI	39.579^***	−2.152	−3.431^***

Note: ***, **, and * are significant at 1%, 5%, and 10%, respectively. ∆ represents the first-order difference of the variable.

In order to ensure the long-run equilibrium between variables, we further conduct cointegration test based on the panel unit root test. As mentioned above, all variables are integrated of order 1, indicating that our data conforms to the premise of cointegration test. In this paper, the Kao test method is adopted for cointegration test, and its null hypothesis is that there is no cointegration relationship. It can be seen from Table 4 that all the five test statistics reject the null hypothesis at least at the significance level of 5%, indicating that there is a long-run stable cointegration relationship among variables. Therefore, all variables in this paper can be used to estimate parameters through the panel data model.

Table 4.

Results of cointegration test.

Test method	Statistic	Statistic value	p Value
Kao test	Modified Dicky-Fuller t	−1.8993	0.0288
	Dicky-fuller t	−3.9896	0.0000
	Augmented Dicky-Fuller t	−2.6242	0.0043
	Unadjusted modified Dicky-Fuller t	−1.8615	0.0313
	Unadjusted Dicky-Fuller t	−3.9686	0.0000

Note: The procedure “demean” is specified for each Kao test to mitigate the impact of cross-sectional dependence.

Analysis of results

The paper tests the models for heteroscedasticity and autocorrelation before regression. The results are shown in table 5. All tests significantly reject the null hypothesis, indicating that there is heteroskedasticity and autocorrelation in the models. Therefore, in order to correct the heteroskedasticity and autocorrelation problems, all regressions in this paper use cluster robust standard errors and cluster at the city level. The empirical results are shown in Table 6.

Table 5.

Results of heteroscedasticity and autocorrelation.

Models	Test method	Statistic	Statistic value
Model between tourism specialization and economic growth	Heteroskedasticity	Modified Wald test	1349.17
Model between tourism specialization and economic growth	Autocorrelation	Wooldridge test	375.390
Model between related diversity and economic growth	Heteroskedasticity	Modified Wald test	1386.46
Model between related diversity and economic growth	Autocorrelation	Wooldridge test	524.759
Model between unrelated diversity and economic growth	Heteroskedasticity	Modified Wald test	1220.41
Model between unrelated diversity and economic growth	Autocorrelation	Wooldridge test	530.100

Table 6.

Regression results of threshold model.

Variable	(1)	(2)	(3)
lnTR_0	0.359^***
(lnPOP ≤ γ ₀ )	(0.060)
lnTR_1	0.460^***
(γ ₀ < lnPOP < γ ₁ )	(0.065)
lnTR_2	0.641^***
(lnPOP ≥ γ ₁ )	(0.066)
lnRV		0.281^*
lnRV		(0.152)
lnUV_0			−0.612^**
(lnPOP ≤ γ ₂ )			(0.246)
lnUV_1			0.491^**
(lnPOP > γ₂)			(0.221)
lnEDU	0.285^***	0.326^***	0.337^***
lnEDU	(0.084)	(0.111)	(0.109)
lnINF	0.253^***	0.466^***	0.455^***
lnINF	(0.084)	(0.102)	(0.098)
InIND	0.066^**	0.024	0.045
InIND	(0.034)	(0.041)	(0.039)
lnGOV	0.307^***	0.467^***	0.458^***
lnGOV	(0.666)	(0.097)	(0.094)
lnFDI	−0.043^**	−0.064^***	−0.062^***
lnFDI	(0.016)	(0.018)	(0.018)
Constant	6.458^***	6.719^***	6.686^***
Constant	(0.322)	(0.454)	(0.461)
γ₀	6.6969^*
γ₁	7.6350^*
γ₂			6.6481^**
F	108.26^***	80.55^***	108.79^***
R²	0.7983	0.6971	0.7146
N	900	900	900

Notes: γ₀、γ₁、and γ₂ are threshold values. The values above the brackets are the estimated coefficients of the variables, and the values in the brackets are the corresponding cluster robust standard errors. ***, **, and * denote significant at the 1%, 5%, and 10% levels, respectively.

Column (1) of Table 6 reports the estimation results of tourism specialization on economic growth of tourist cities based on the threshold model. In column (1), there are two thresholds for the impact of tourism specialization on economic growth of tourist cities, which are 6.6969 (the city’s resident population is 8.0986 million) and 7.6350 (the city’s resident population is 20.693 million), respectively. When the city size is lower than the first threshold value, the impact of tourism specialization on city’s economic growth is significantly positive, and it passes the 1% significance level test with a coefficient value of 0.359. When the city size exceeds the first threshold value, the impact of tourism specialization on city’s economic growth increases, with a coefficient of 0.460 and passing the 1% significance level test. Furthermore, when the city size exceeds the second threshold value, the impact of tourism specialization on city’s economic growth is further enhanced, and the coefficient rises to 0.641, which still passes the 1% significance level test. The above analysis results show that the impact of tourism specialization level on economic growth of tourist cities is related to city size. Specifically, tourism specialization has a positive impact on economic growth of tourist cities, and there is a double threshold effect based on city size. When the city size exceeds the two thresholds, the promotion effect of tourism specialization on economic growth of tourist cities has a significant jump. This is consistent with the research results of Huang et al. (2000) and Zhang and Sun (2016) and verifies our H1. Tourism specialization is more advantageous in large-scale tourist cities. This is mainly because tourist cities with larger city size tend to have more advantages in infrastructure system, tourism planning and development capacity, consumer market, and other aspects. As a result, large cities are more attractive to tourists, which is conducive to the aggregation of advantageous resources and element allocation.

According to the above analysis, there is no threshold effect between RV and economic growth of tourist cities based on city size. Therefore, we use a linear regression model to test their relationship. Column (2) of Table 5 is the result of parametric regression. In order to achieve better fitting effect, it is still necessary to determine whether the model is suitable for mixed regression model, fixed effect model, or random effect model before panel model regression. After F test and Hausman test on the model, it is found that the null hypothesis of the mixed effect model and random effect model are rejected at the 1% significance level. Thus, the fixed effect model is finally selected for analysis in this paper. In column (2), the coefficient of industry-related diversity is significantly positive at the 5% level, indicating that the RV has a significant positive effect on economic growth of tourist cities. Related diversity emphasizes the aggregation of industries, which makes knowledge spillovers more likely to occur, so as to promote economic and technological connections between local enterprises. This also reduces the enterprises’ cost, thereby improving the enterprises’ productivity, and may also achieve common economic development by attracting related industries to gather.

Column (3) of Table 6 reflects the parametric regression results of UV on economic growth of tourist cities. The threshold value of UV to economic growth of tourist cities is 6.6481 (the resident population of the city is 7.713 million). When the city size is less than the threshold value, UV inhibits city’s economic growth with a coefficient of −0.612 and is tested at the 5% significance level. When the city size expands and crosses the threshold value, the further improvement of the level of UV will promote economic growth of tourist cities. It has a coefficient of 0.491, which still passes the 5% significance level. Therefore, there is a significant U-shaped threshold effect based on city size in the impact of UV on economic growth of tourist cities. This is consistent with H2. With the expansion of city size, diversified industrial structure is concentrated in the region, forming diversified industrial spatial distribution pattern. This can strengthen the close linkages between the various industries, resulting in a wider range of local economic benefits. Compared with small cities, the diversity of large cities is more conducive to the exchange of information, sharing of knowledge and technology, thereby promoting economic growth of tourist cities.

For control variables, the estimated coefficients of human capital, infrastructure and government intervention on economic development are much higher than the degree of openness and industrial structure. It shows that the current economic growth of tourist cities still relies more on the accumulation of human capital, the improvement of infrastructure, and the investment of government finance in China. However, the degree of openness has a significant hindering effect on the economic development of tourist cities, which may be due to the low degree of openness and the large differences in the regional distribution of the degree of utilizing foreign investment in these cities. Compared with the central and western regions, the eastern coastal regions are more likely to expand the level of opening up and improve the quality of economic growth.

Robustness test

This paper uses two methods to test the robustness of the empirical results. On the one hand, all continuous variables are winsorized at the 1% and 99% levels to prevent extreme values from adversely affecting the estimated results. On the other hand, we change the measurement index of the dependent variable, use real GDP (lnGDP) to represent economic growth level of tourist cities, and re-estimate the model parameters.

First, the robustness test results of the threshold effect of city size are shown in Table 7. Panel A is the result after being winsorized at the 1% and 99% levels of the data, and panel B is the result after replacing the dependent variable. These results show that the single threshold and double threshold of tourism specialization are significant at the 10% level. However, the triple threshold effect does not pass the significance test, indicating that tourism specialization has a double threshold effect on economic growth of tourist cities. Related diversity has no threshold effect of city size on economic growth of tourist cities, while UV has a single threshold effect on economic growth of tourist cities. The threshold values and their confidence intervals are also close to the original results. It can be seen that the test results of threshold effect of city size are relatively robust.

Table 7.

Robustness test results of threshold effect of city size.

Key dependent variables	Threshold type	F stat	Critical value			Threshold value	Confidence interval
Key dependent variables	Threshold type	F stat	10%	5%	1%	Threshold value	Confidence interval
	A. Being winsorized at the 1% and 99% levels
Tourism specialization	Single	52.16^*	45.89	57.94	85.80	7.6102^*	(6.6713, 7.6350)
	Double	53.97^*	30.31	56.47	154.43	7.6102^*	(7.5817, 7.6350)
	Double					6.6481^*	(6.6384, 6.6542)
	Triple	28.38	54.38	70.19	154.81
Related diversity	Single	11.81	23.04	27.37	51.06
Unrelated diversity	Single	60.45^**	42.54	50.18	60.30	6.6481^**	(6.6403, 6.6542)
	Double	34.85	38.87	45.29	62.63
	Triple	22.47	59.43	104.83	143.37
	B. Replacing the dependent variable
Tourism specialization	Single	93.34^***	57.34	66.39	78.59	6.6481^***	(6.6384, 6.6542)
	Double	64.23^**	48.18	52.76	68.04	6.6481^**	(6.6384, 6.6542)
	Double	64.23^**	48.18	52.76	68.04	7.6102^**	(7.5817, 7.6350)
	Triple	48.35	76.60	99.18	169.66
Related diversity	Single	15.66	25.78	31.04	43.11
Unrelated diversity	Single	85.05^***	42.35	47.97	62.27	6.6481^***	(6.6412, 6.6542)
	Double	19.84	29.09	32.57	45.07
	Triple	17.84	39.18	48.89	61.84

Second, the robustness test results of the threshold model regression are shown in Table 8. Columns (1)–(3) are the results after being winsorized at the 1% and 99% levels of the data (the robustness test A). In column (1), the coefficients of lnTR_0, lnTR_1, and lnTR_2 are significantly positive at the level of 1% and increase successively, indicating that tourism specialization has a positive impact on economic growth of tourist cities, and the promotion effect of tourism specialization on economic growth of tourist cities has a significant jump when the city size exceeds the two thresholds. In column (2), the coefficient of ln RV is significantly positive at the level of 5%, showing that the RV can significantly promote economic growth of tourist cities. In column (3), the coefficient of lnUV_0 is negative and the coefficient of lnUV_1 is significantly positive at the level of 1%, demonstrating that UV will promote economic growth of tourist cities when the city size is higher than the threshold value. In general, the results above are consistent with the original results. Therefore, the robustness test A can confirm the robustness of the conclusions in this paper.

Table 8.

Robustness test results of threshold model regression.

	A. Being winsorized at the 1% and 99% levels			B. Replacing the dependent variable
Variables	(1)	(2)	(3)	(4)	(5)	(6)
lnTR_0	0.302^***			0.374^***
(lnPOP ≤ γ ₀ )	(0.053)			(0.069)
lnTR_1	0.390^***			0.525^***
(γ ₀ < lnPOP < γ ₁ )	(0.055)			(0.074)
lnTR_2	0.579^***			0.783^***
(lnPOP ≥ γ ₁ )	(0.061)			(0.077)
ln RV		0.357^**			0.355^*
		(0.141)			(0.192)
lnUV_0			−0.381			−0.775^**
(lnPOP ≤ γ ₂ )			(0.259)			(0.340)
lnUV_1			0.636^***			0.683^**
(lnPOP > γ₂)			(0.222)			(0.317)
lnEDU	0.254^***	0.276^***	0.292^***	0.263^***	0.300^*	0.314^***
	(0.080)	(0.097)	(0.096)	(0.084)	(0.110)	(0.106)
lnINF	0.326^***	0.533^***	0.511^***	0.286^***	0.517^***	0.500^***
	(0.069)	(0.067)	(0.065)	(0.089)	(0.110)	(0.105)
lnIND	0.056^*	0.019	0.038	0.076^*	0.023	0.052
	(0.032)	(0.038)	(0.036)	(0.042)	(0.050)	(0.047)
lnGOV	0.399^***	0.549^***	0.544^***	0.371^***	0.549^***	0.536^***
	(0.058)	(0.083)	(0.080)	(0.082)	(0.114)	(0.110)
lnFDI	−0.039^**	−0.050^**	−0.049^**	−0.066^***	−0.090^***	−0.087^***
	(0.015)	(0.017)	(0.016)	(0.020)	(0.022)	(0.021)
Constant	6.475^***	6.696^***	6.593^***	3.257^***	3.608^***	3.556^***
	(0.309)	(0.400)	(0.408)	(0.324)	(0.444)	(0.455)
γ₀	6.6481^**			6.6481^**
γ₁	7.6102^**			7.6102^**
γ₂			6.6481^**			6.6481^***
F	108.59^***	130.14^***	149.22^***	98.40^***	72.65^***	90.71^***
R²	0.8175	0.7488	0.7613	0.7791	0.6723	0.6966
N	900	900	900	900	900	900

Notes: γ₀, γ₁, and γ₂ are threshold values. The values above the brackets are the estimated coefficients of the variables, and the values in the brackets are the corresponding cluster robust standard errors. ***, **, and * denote significant at the 1%, 5%, and 10% levels, respectively.

Moreover, columns (4)–(6) are the results after replacing the dependent variable (the robustness test B). In column (4), the signs and magnitudes of the coefficients of lnTR_0, lnTR_1, and lnTR_2 are consistent with the original results and the robustness test A. In column (5), the coefficient of ln RV is significantly positive at the level of 10%, which is also consistent with the original results and the robustness test A. In column (6), the coefficient of lnUV_0 is significantly negative at the level 5% and the coefficient of lnUV_1 is significantly positive at the level of 5%, indicating that there is a significant U-shaped threshold effect based on city size in the impact of UV on economic growth of tourist cities. This further verifies the original results. To sum up, the robustness test B show that the signs and magnitudes of the coefficients are basically consistent with the original results, which further confirms the robustness of the conclusion.

Conclusions and implications

Main conclusions

From the perspective of city size, this paper innovatively studies the role of diversity and specialization on economic growth of tourist cities of different sizes, aiming to provide basis for the formulation of industrial development strategy and pattern design of tourist cities. The findings are as follows: first, the promotion effect of tourism specialization on economic growth of tourist cities is related to city size. When the city size exceeds the two thresholds, the promotion effect of tourism specialization on economic growth of tourist cities has obvious jump. Second, the promotion effect of RV on economic growth of tourist cities has nothing to do with city size. Third, the impact of UV on economic growth of tourist cities presents a U-shape. When the city size is lower than the threshold value, UV inhibits economic growth of tourist cities. However, when the city size exceeds the threshold value, UV promotes economic growth of tourist cities.

Theoretical implications

Based on the empirical findings, the paper can formulate two theoretical implications for research on agglomeration externalities and economic growth. First, the results offer evidence for the nonlinear effects of tourism specialization on economic growth of tourist cities from the perspective of city size, which contributes to the tourism agglomeration externalities theory. The impact of city size has been neglected in previous studies on the relationship between tourism development and economic growth. This paper uses the threshold regression model with city size as the threshold variable and finds that tourism specialization has a positive impact on economic growth of tourist cities, and there is a double threshold effect based on city size. Second, this paper puts diversified externalities into the research frame of economic growth of tourist cities, which enriches the theory of diversification externalities. Prior studies have mainly focused on the impact of tourism specialization on economic growth of tourist cities, ignoring the role of industrial diversity. This paper uses entropy method to decompose industrial diversity into two parts, related and unrelated, and discusses their impact on economic growth of tourist cities.

Policy implications

The main policy implications of this paper are as follows: tourist cities should reasonably design and adjust the industrial structure according to the city size, and avoid blindly pursuing the development of tourism specialization or industrial diversity. In other words, tourist cities should first consider their own population when formulating economic development strategies and then carry out industrial economic activities. Specifically, first of all, for small- and medium-sized tourist cities, they should concentrate superior resources to develop tourism and give full play to MAR externalities and advantages of localized economy. At the same time, according to the comparative advantages of the region and the correlation between the characteristics of the industry, the communication and integration of tourism industry and other related industries should be promoted to highlight economic growth effect of RV. Second, for large tourist cities, on the premise of realizing scale economy in the existing tourism industry, they should rely on the macro-control function of the government to guide the diversified adjustment of the industrial structure, so that the stabilizer function of UV can be fully played. The synergistic development of various industries should be encouraged to give play to Jacobs externalities among industries. Based on the dominant industry status of tourism, these cities should continue to develop industries that have a certain foundation and are horizontally related to the tourism industry, and improve the level of RV. In combination with the strategic focus of the national strategic emerging industries, tourist cities should appropriately cultivate emerging sectors that are different from the existing pillar industries against the economic cycle and improve the level of UV.

Limitations and future research directions

Although this study provides several significant implications, there are still certain limitations that need to be addressed in future research. First, considering the availability of data, the paper is limited to a single geographic area. Although tourist cities in China are studied because of the current importance of tourism development in China, future research can collect corresponding data from other countries, therefore resolving the generalizability issue. Second, due to space constraints, the paper only uses city size as the threshold variable, and examines the nonlinear effects of specialization and diversity from the perspective of city size. Future research could establish other threshold variables to better understand the nonlinear effect of tourism specialization and diversity.

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 sponsored by Shanghai Pujiang Program (NO.21PJC054).

ORCID iDs

Weishu Zhao

Guanxu Wan

Taotao Deng

Author biographies

Mulan Ma is an associate professor in Shanghai University of International Business and Economics. She received her PhD from University of Kiel, Germany. Her research interests include tourism management and tourism economics. Her publications have appeared in Annals of Tourism Research, Journal of Travel Research, Tourism Geographies, and Tourism Economics.

Weishu Zhao is a PhD candidate in Shanghai University of Finance and Economics. Her research interests include tourism management and regional economics. Her publications have appeared in Journal of Travel Research, Tourism Economics, and Business Management Journal.

Guanxu Wan is a PhD candidate in Shanghai University of Finance and Economics. His research interests include tourism management and regional economics. His publications have appeared in International Journal of Technology Management.

Taotao Deng is a professor in Shanghai University of Finance and Economics. He received his PhD from University of Aberdeen, UK. His research interests lie in regional studies, tourism economics, and transport economics. As the lead author, he has authored 26 scholarly papers indexed by SSCI. His publications have appeared in Annals of Tourism Research, Journal of Travel Research, Tourism Economics, Transportation, Transport Reviews, and Cities.

Yi Yang is a postgraduate in Shanghai University of International Business and Economics. Her research interests include tourism management and tourism economics.

References

Aarstad

Kvitastein

Jakobsen

S-E

(2016) Related and unrelated variety in a tourism context. Annals of Tourism Research 57: 254–256.

Adamou

Clerides

(2010) Prospects and limits of tourism-led growth: the international evidence. Review of Economic Analysis 2(3): 287–303.

Arrow

(1962) The economic implications of learning by doing. The Review of Economic Studies 29(3): 155–173.

Arthur

(1920) The Economics of Welfare. London: Macmillan Company Inc.

Attaran

(1986) Industrial diversity and economic performance in US areas. The Annals of Regional Science 20(2): 44–54.

Batisse

(2002) Dynamic externalities and local growth: a panel data analysis applied to Chinese provinces. China Economic Review 13(2–3): 231–251.

Boschma

(2005) Proximity and innovation: a critical assessment. Regional Studies 39(1): 61–74.

Brau

Lanza

Pigliaru

(2007) How fast are small tourism countries growing? Evidence from the data for 1980–2003. Tourism Economics 13(4): 603–613.

Brida

Giuliani

(2013) Empirical assessment of the tourism-led growth hypothesis: the case of the Tirol-Südtirol-Trentino. Tourism Economics 19(4): 745–760.

10.

Browning

Singelman

(1975) The Emergence of a Service Society: Demographic and Sociological Aspects of the Sectoral Transformation of the Labor Force in the USA. Beverly Hilly, CA: Sage Publication.

11.

Chang

Khamkaew

Mcaleer

(2012) IV estimation of a panel threshold model of tourism specialization and economic development. Tourism Economics 18(1): 5–41.

12.

Croes

Dr (2013) Tourism specialization and economic output in small islands. Tourism Review 68(4): 34–48.

13.

Croes

Ridderstaat

Niekerk

(2018) Connecting quality of life, tourism specialization, and economic growth in small island destinations: the case of Malta. Tourism Management 65: 212–223.

14.

Croes

Ridderstaat

Bąk

, et al. (2021) Tourism specialization, economic growth, human development and transition economies: the case of Poland. Tourism Management 82: 104181.

15.

De Vita

Kyaw

(2017) Tourism specialization, absorptive capacity, and economic growth. Journal of Travel Research 56(4): 423–435.

16.

Deng

Cao

(2014) Tourism resource development and long-term economic growth: a resource curse hypothesis approach. Tourism Economics 20(5): 923–938.

17.

Deng

Shao

(2014) Has international tourism promoted economic growth in China? A panel threshold regression approach. Tourism Economics 20(4): 911–917.

18.

Dissart

(2003) Regional economic diversity and regional economic stability: research results and agenda. International Regional Science Review 26(4): 423–446.

19.

Dogru

Marchio

Bulut

, et al. (2019) Climate change: vulnerability and resilience of tourism and the entire economy. Tourism Management 72: 292–305.

20.

Duranton

Puga

(2005) From sectoral to functional urban specialisation. Journal of Urban Economics 57(2): 343–370.

21.

Duranton Puga Diego (2001) Nursery cities: urban diversity, process innovation, and the life cycle of products. The American Economic Review 91(5): 1454–1477.Gilles.

22.

Duro

Perez-Laborda

Turrion-Prats

, et al. (2021) Covid-19 and tourism vulnerability. Tourism Management Perspectives 38(2): 100819.

23.

Erkuş-Öztürk

Terhorst

(2018) Economic diversity of a single-asset tourism city: evidence from Antalya. Current Issues in Tourism 21(4): 422–439.

24.

Fan

Feng

(2014) Industrial agglomeration and total factor productivity of enterprises. Journal of World Economics 37(5): 51–73. (in Chinese).

25.

Feldman

Audretsch

(1999) Innovation in cities: science-based diversity, specialization and localized competition. European Economic Review 43(2): 409–429.

26.

Frenken

Van Oort

Verburg

(2007) Related variety, unrelated variety and regional economic growth. Regional Studies 41(5): 685–697.

27.

Fritsch

Kublina

(2018) Related variety, unrelated variety and regional growth: the role of absorptive capacity and entrepreneurship. Regional Studies 52(10): 1360–1371.

28.

Gil-Alana

Poza

(2020) The impact of covid-19 on the Spanish tourism sector. Tourism Economics 28: 646–653.

29.

Glaeser

Kallal

Scheinkman

, et al. (1992) Growth in cities. Journal of Political Economy 100(6): 1126–1152.

30.

Gssling

Scott

Hall

(2021) Pandemics, tourism and global change: a rapid assessment of covid-19. Journal of Sustainable Tourism 29(1): 1–20.

31.

Haugland

Ness

Grønseth

B-O

, et al. (2011) Development of tourism destinations: an integrated multilevel perspective. Annals of Tourism Research 38(1): 268–290.

32.

Henderson

(1986) Efficiency of resource usage and city size. Journal of Urban Economics 19(1): 47–70.

33.

Henderson

Kuncoro

Turner

(1995) Industrial development in cities. Journal of Political Economy 103(5): 1067–1090.

34.

Huang

Hou

(2000) Preliminary probing into tourism urbanization: a case study of metropolitan interlocking region in Yangtze River. Resources and Environment in the Yangtze Basin 9(2): 160–165. (in Chinese).

35.

Jackman

(2014) Output volatility and tourism specialization in small island developing states. Tourism Economics 20(3): 527–544.

36.

Jacobs

(1969) The Economy of Cities. New York: Vintage.

37.

Jacobs

(1984) Cities and the Wealth of Nation: Principles of Economic Life. New York: Vintage.

38.

Jin

(2011) The effects of tourism on economic growth in Hong Kong. Cornell Hospitality Quarterly 52(3): 333–340.

39.

Zhao

(2014) Industrial structure, city size and urban productivity in China. Economic Research Journal 49(4): 76–88+115. (in Chinese).

40.

Kreiner

Ram

(2020) National tourism strategies during the covid-19 pandemic. Annals of Tourism Research 89: 103076.

41.

Krugman

(1991) Increasing returns and economic geography. Journal of Political Economy 99(3): 484–499.

42.

Lanza

Temple

Urga

(2003) The implications of tourism specialisation in the long run: an econometric analysis for 13 OECD economies. Tourism Management 24(3): 315–321.

43.

Lee

(2012) Tourism, trade, and income: evidence from Singapore. Anatolia 23(3): 348–358.

44.

Wang

(2014) The four worlds of Chinese social stratification. Social Science Front 37(9): 174–187. (in Chinese).

45.

Lim

(2021) The economic impact of a global pandemic on the tourism economy: the case of COVID-19 and Macao’s destination- and gambling-dependent economy. Current Issues in Tourism 25(8): 1258.

46.

Liu

Pratt

(2017) Tourism’s vulnerability and resilience to terrorism. Tourism Management 60: 404–417.

47.

Marshall

(1920) Principles of Economics. London: Macmillan.

48.

Marsiglio

(2018) On the implications of tourism specialization and structural change in tourism destinations. Tourism Economics 24(8): 945–962.

49.

Neffke

Henning

Boschma

, et al. (2011) The dynamics of agglomeration externalities along the life cycle of industries. Regional Studies 45(1): 49–65.

50.

Neuts

(2020) Tourism and urban economic growth: a panel analysis of German cities. Tourism Economics 26(3): 519–527.

51.

Pesaran

(2004) General diagnostic test for cross section dependence in panels. In: Cambridge Working Papers in Economics. No. 0435, Faculty of Economics: University of Cambridge.

52.

Pesaran

(2007) A Simple Panel Unit Root Test in the Presence of Cross- Sectional Dependence. Journal of Applied Econometrics 22(2): 265–312.

53.

Romer

(1986) Increasing returns and long-run growth. Journal of Political Economy 94(5): 1002–1037.

54.

Romer

(1990) Endogenous technological change. Journal of Political Economy 98(5): 71–102.

55.

Sequeira

Nunes

(2008) Does tourism influence economic growth? A dynamic panel data approach. Applied Economics 40(18): 2431–2441.

56.

Sheng

(2011) Specialisation versus diversification: a simple model for tourist cities. Tourism Management 32(5): 1229–1231.

57.

Siegel

Johnson

Alwang

(1995) Regional economic diversity and diversification. Growth and Change 26(2): 261–284.

58.

Simonen

Svento

Juutinen

(2015) Specialization and diversity as drivers of economic growth: evidence from high-tech industries. Papers in Regional Science 94(2): 229–247.

59.

Skyes

(1950) Diversification of industry. The Economic Journal 60(240): 697–714.

60.

Soulie

Valle

(2014) Trade effects of specialization in tourism: an inter-regional input–output model of the balearic islands. Tourism Economics 20(5): 961–985.

61.

Zhao

(2011) Sectoral specialization, functional specialization and growth in cities: evidences from panel data on China’s prefecture-level city. China Industrial Economic 27(4): 25–34. (in Chinese).

62.

Sun

Chai

(2012) Related diversification, unrelated diversification and regional economic development: evidences from panel data on China’s 282 prefecture-level city. China Industrial Economic 28(6): 5–17. (in Chinese).

63.

Sun

Zhou

(2013) Diversification, specialization, city size and economic growth: an empirical analysis based on panel data of prefecture-level cities of China. Journal of Industrial Engineering and Engineering Management 27(2): 71–78. (in Chinese).

64.

Van Oort

De Geus

Dogaru

(2015) Related variety and regional economic growth in a cross-section of european urban regions. European Planning Studies 23(6): 1110–1127.

65.

Wagner

Deller

(1998) Measuring the effects of economic diversity on growth and stability. Land Economics 74(4): 541–556.

66.

Wang

Xia

(2013) A study on the relationship between tourism industry and regional economic growth – a case study of Jiangsu Gaochun district. Modern Economy 4: 482–488.

67.

Zhang

(2020) The influence of high-speed railway on the development of urban tourism economy based on the PSM-DID model. Journal of Huaqiao University 38(5): 53–64. (in Chinese).

68.

Zhang

Sun

(2016) Spatial relation between city size and tourism growth, and its evolution and driving factors: new understanding of city level system in the yangtze river delta. Scientia Geographica Sinica 36(12): 1877–1884.

69.

Zhang

Zhou

, et al. (2020) Higher tourism specialization, better hotel industry efficiency? International Journal of Hospitality Management 87: 102509.

70.

Zhang

Song

Wen

, et al. (2021) Forecasting tourism recovery amid covid-19. Annals of Tourism Research 87: 103149. (in Chinese).