**3. Methodology**

*Peripheral Territories, Tourism, and Regional Development*

should be the focus of public policies and strategies.

and Surugiu [26] studied the relationship between the tourism sector and economic growth in Romania from 1988 to 2009 using econometric cointegration Granger causality methods, vector error correction model (VECM), and impulse response functions. The findings show that tourism expansion does granger cause economic growth. Results from this study place a focus on the requirement for effective tourism development strategies. Fundeanu [27] analysed the role of tourism clusters in the south-west Oltenia region, looking at tourism potential, tourism diversity, strengths, and weaknesses of the region. The study found that tourism clusters are catalysts for regional development, and the competitive advantages of such clusters

Gunderson and Ng [28] studied the impact of tourism on regional development in the rural USA. Tourist spending could result in increased demand for regional goods and services, eventually leading to employment creation and an increase in disposable income. The results indicate that public policy effectiveness, sustainable natural resource management, and community development could allow for tourism development and regional development. Tourism positively affects regional economic performance. Klytchnikova and Dorosh [29] analysed the role of tourism on regional economic development in Panama's poor regions using a social accounting matrix model. The paper used the impact of tourism spending on growth and poverty at the regional level. The results indicate that tourism has a large impact on the regional economy and is also an important multiplier in the local economy. The sector also allows for important benefits to the poor. Mishra, Rout and Mohapatra [30] considered tourism an important sector to promote regional economic growth and analysed the import of India's sector from 1978 to 2009 using econometric methods. Time-series econometric models were used for the analysis from 1978 to 2009. The results indicate the existence of unidirectional causality running from tourism activities to economic growth. Wen-li [31] analysed the impact of tourism on economic growth in regional China since the 1990s. The study results indicate a significant impact of tourism on regional economic growth and allow for diversified and balanced development. He and Zheng [32] analysed the Sichuan region from 1990 to 2009 in China and the tourism sector's impact with abundant resources on the regional economy. Over the last decade, the contribution of tourism to the provincial GDP has been increasing annually. Results indicated that a bi-directional relationship exists between tourism development and economic growth. Yang, Fik, and Altschuler [33] analysed tourism-related economic multipliers from regional input–output tables for 30 Chinese provinces looking at tourism variables, including income, employment, and employment multipliers. Interesting findings reveal that the output and employment multipliers of tourism are positively associated with regional economic development. Rogerson [34] states that uneven development is a reality of South Africa's spatial economy's structure with leading and lagging regions. Tourism has been identified as a vital economic sector for regional development. This paper assesses the 23 distressed regions in dire need of economic development. These regions rely mostly on domestic tourism, and local natural assets should be maximised with effective policy implementation. Meyer and Meyer [8] conducted a study using regional tourism statistical data from 2001 to 2013 for two geographical areas in a developing region in South Africa. The results indicate that tourism in these regions has a significant impact on economic growth as the sector does include low skilled workers in a labour-intensive industry and allows for a range of benefits for regions

Lastly, the tourism sector could also have negative impacts, especially on the environment and sustainable development. Effective policies should be in place to allow for strategies to prevent the environment's deterioration [35]. Pedrana [11] believes that tourists could negatively affect local cultures. These negative impacts could be

**106**

that include employment and income.

The research methodology is based on a quantitative analysis using both descriptive and econometric methods to achieve the primary objective. In this study, the tourism sector's impact on a developing region is analysed using secondary data from Global Insight [37]. Annual data from 1996 to 2019 were used and analysed firstly utilising trends and correlation analysis and secondly by using a pooled econometric panel approach including the five municipal areas (see **Figure 1** for details) in the Gauteng province in South Africa. Gauteng province comprises of the following municipal regions:


The variables used in the panel econometric analysis consisted of GDP per capita representing economic development and growth as the dependent variable, with the following independent variables: gross value added (GVA) in the tourism sector; the number of jobs in the tourism sector; spending per capita in the tourism sector; and the number of international trips to the region. The panel data for the Gauteng region were analysed using a multiple regression. A multiple regression includes many variables to predict changes in the dependent variable [38]. All of the variables were converted into natural logarithms. The variables were set as follows with the abbreviations as used:

Dependent variable (Y) = Economic development and growth (GDPC - the log format LGDPC)

Independent variables (X):


Pedroni [39] formulated Eq. (1) represents the basic model for the panel data analysis:

$$Y\_{it} = \mathfrak{a}\_i + \delta\_i t + \beta\_i X\_{it} + \mathfrak{e}\_{it} \tag{1}$$

Where:

*Y*it = Dependent variable.

*a*i = Intercept term.

*δi* = Parameter that, together with *a*i allows the individual linear trends and individual effects to be observed respectively [6].

β = k × 1 vector of parameters that were estimated based on the explanatory variables.

*X*it = 1 × k vector of observations of the explanatory variables, t = 1,. .., T; i = 1,. . .

In this econometric analysis process, several models were used to test for long and short-run relationships between variables: (1) unit root tests to determine the level of stationarity of the variables and model selection; (2) long-run relationships between the variables using either a panel ARDL of Fisher-Johansen cointegration test leading to regression analysis using FMOLS and DOLS equations; (3) Granger causality test to assess causality between all the variables; (4) and model stability diagnostic tests. To simplify the analysis of results from all tests and to place all variables on the same scale, all variables were converted into the natural logarithm format. A panel data set was created for the five municipal regions within the Gauteng Province with 120 observations. Eq. (2) represents the basic equation for pooled panel data econometric models, as proposed by Brooks [40]:

$$\text{If } y\_{\text{it}} = \alpha + \beta \mathbf{x}\_{\text{it}} + u\_{\text{it}} \text{ and } I = \mathbf{1}, \dots, N; t = \mathbf{1}, \dots, T \tag{2}$$

**109**

*An Assessment of the Impact of the Tourism Sector on Regional Economic Development…*

1 1 1 1 1

*j j*

+ ++

β

β *<sup>n</sup>* , λ

= =

*LSPENDT LINTTT u*

*k k*

∑ ∑

observations on the explanatory variables, *t* = 1,.., *T*; *i* = 1, i denotes countries in the panel and t denotes time dimension*.* The i subscript denotes the cross-section and t the time series. The model from the function described in Eq. (1) can be listed as

*t j tj j t j t j*

 λ

*t j tj t*

*LGDPC LGDPC LGVAT LJOBST*

=∝ + + +

− −

and *u*<sup>1</sup>*t* and *u*<sup>2</sup>*t* are the stochastic error terms, which are also known as shocks in the model. The following tests were used to determine the stationarity level for all the variables: Im, Pesaran and Shin W-stat test; the ADF - Fisher Chi-square test; and the PP - Fisher Chi-square test. Model selection was based on the unit root results. Stationarity of all variables at levels or I(0), a panel VAR analysis would be estimated. In contrast, if all variables were stationary at 1st difference or I(1), the Fisher Johansen panel cointegration test for long-run relationships should be estimated. Lastly, if a mixture of variables were presented, the optimal option

As mentioned earlier, the focus area selected for this study is the Gauteng Province in South Africa. This region was chosen due to the following reasons. Firstly, it is the economic hub in South Africa and the African continent; secondly, it is rich in cultural and historical places and events. The region is the largest metropolitan region in the world not located adjacent to the ocean or a major water body. Based on the aforementioned information, this research study's main objective is to assess the dynamic economic impact of tourism on the Gauteng province, using

In terms of the descriptive analysis, eight key variables were selected as indicated in **Table 1**, to compare the five municipal regions with the total Gauteng province. Variables were analysed regarding growth rates, trends, and contributions to the study region. The different descriptive variables are analysed individually. Firstly, the GDP at constant prices is analysed. The Gauteng province had strong economic growth from 2009 to 2014 of 3.1% per annum, but growth has been low and slow from 2014 to 2019 at 1.1%. The COJ contributed the most to the provincial GDP of 44%, followed by the COT. The two peripheral regions of SDM and WRDM only contributed 4% and 3.8% to the province. The COT had the highest annual growth rate of 1.3% from 2014 to 2019. Secondly, Gauteng has a declining situation regarding GDP per capita with a negative growth rate of −1.2% from 2014 to 2019. Of the five municipal areas, both COJ and COT had much higher GDP per capita values and negative growth rates. SDM had the lowest GDP per capita at less than half of the two leading regions, namely COJ and COT. Thirdly, in terms of population density, Gauteng province had a density of 649 per sqkm and has an increased rate of 2.8% per annum. COJ and EKR metros had the highest densities of 2630 and 1562, with high levels of increases. The two more rural regions of SDM and WRDM have much lower at 208 and 176,

1

− −−

*<sup>n</sup>* are the coefficients, K is the number of lags,

(3)

*DOI: http://dx.doi.org/10.5772/intechopen.95810*

*n* is the constant,

would be a panel ARDL method as an estimation.

both descriptive and time-series approaches.

follows in Eq. (3):

Where

α

**4. Results and discussion**

**4.1 Descriptive analysis**

Where *y*it is the dependent variable, *α* is the intercept term, *β* is a *k* × 1 vector of parameters to be estimated on the explanatory variables, and *x*it is a 1 × *k* vector of

*An Assessment of the Impact of the Tourism Sector on Regional Economic Development… DOI: http://dx.doi.org/10.5772/intechopen.95810*

observations on the explanatory variables, *t* = 1,.., *T*; *i* = 1, i denotes countries in the panel and t denotes time dimension*.* The i subscript denotes the cross-section and t the time series. The model from the function described in Eq. (1) can be listed as follows in Eq. (3):

$$\begin{aligned} LGDPC\_t &= \alpha\_{1} + \sum\_{j=1}^{k} \beta\_{1j} LGDPC\_{t-j} + \sum\_{j=1}^{k} \lambda\_{1j} LGDVAT\_{t-j} + LJOBST\_{t-j} \\ &+ LSPENDT\_{t-j} + LINTTT\_{t-j} + u\_{1t} \end{aligned} \tag{3}$$

Where α*n* is the constant, β *<sup>n</sup>* , λ*<sup>n</sup>* are the coefficients, K is the number of lags, and *u*<sup>1</sup>*t* and *u*<sup>2</sup>*t* are the stochastic error terms, which are also known as shocks in the model. The following tests were used to determine the stationarity level for all the variables: Im, Pesaran and Shin W-stat test; the ADF - Fisher Chi-square test; and the PP - Fisher Chi-square test. Model selection was based on the unit root results. Stationarity of all variables at levels or I(0), a panel VAR analysis would be estimated. In contrast, if all variables were stationary at 1st difference or I(1), the Fisher Johansen panel cointegration test for long-run relationships should be estimated. Lastly, if a mixture of variables were presented, the optimal option would be a panel ARDL method as an estimation.

#### **4. Results and discussion**

#### **4.1 Descriptive analysis**

*Peripheral Territories, Tourism, and Regional Development*

with the abbreviations as used:

Independent variables (X):

LSPENDT)

analysis:

Where:

variables.

*Y*it = Dependent variable. *a*i = Intercept term.

individual effects to be observed respectively [6].

The variables used in the panel econometric analysis consisted of GDP per capita representing economic development and growth as the dependent variable, with the following independent variables: gross value added (GVA) in the tourism sector; the number of jobs in the tourism sector; spending per capita in the tourism sector; and the number of international trips to the region. The panel data for the Gauteng region were analysed using a multiple regression. A multiple regression includes many variables to predict changes in the dependent variable [38]. All of the variables were converted into natural logarithms. The variables were set as follows

Dependent variable (Y) = Economic development and growth (GDPC - the log format LGDPC)

• Gross value added in the tourism sector (GVAT – the log format LGVAT)

• Number of jobs in the tourism sector (JOBST – the log format LJOBST)

• Number of international tourist trips (INTTT – the log format LINTTT)

Pedroni [39] formulated Eq. (1) represents the basic model for the panel data

*Ya t Xe it i i i it it* =+ + + δ β

*δi* = Parameter that, together with *a*i allows the individual linear trends and

β = k × 1 vector of parameters that were estimated based on the explanatory

pooled panel data econometric models, as proposed by Brooks [40]:

α β

*X*it = 1 × k vector of observations of the explanatory variables, t = 1,. .., T; i = 1,. . . In this econometric analysis process, several models were used to test for long and short-run relationships between variables: (1) unit root tests to determine the level of stationarity of the variables and model selection; (2) long-run relationships between the variables using either a panel ARDL of Fisher-Johansen cointegration test leading to regression analysis using FMOLS and DOLS equations; (3) Granger causality test to assess causality between all the variables; (4) and model stability diagnostic tests. To simplify the analysis of results from all tests and to place all variables on the same scale, all variables were converted into the natural logarithm format. A panel data set was created for the five municipal regions within the Gauteng Province with 120 observations. Eq. (2) represents the basic equation for

1 ., ; 1, ., *it it it y x u and I N t T* = + + =… = …

Where *y*it is the dependent variable, *α* is the intercept term, *β* is a *k* × 1 vector of parameters to be estimated on the explanatory variables, and *x*it is a 1 × *k* vector of

(2)

(1)

• Spending per capita in the tourism sector (SPENDT – the log format

**108**

As mentioned earlier, the focus area selected for this study is the Gauteng Province in South Africa. This region was chosen due to the following reasons. Firstly, it is the economic hub in South Africa and the African continent; secondly, it is rich in cultural and historical places and events. The region is the largest metropolitan region in the world not located adjacent to the ocean or a major water body. Based on the aforementioned information, this research study's main objective is to assess the dynamic economic impact of tourism on the Gauteng province, using both descriptive and time-series approaches.

In terms of the descriptive analysis, eight key variables were selected as indicated in **Table 1**, to compare the five municipal regions with the total Gauteng province. Variables were analysed regarding growth rates, trends, and contributions to the study region. The different descriptive variables are analysed individually. Firstly, the GDP at constant prices is analysed. The Gauteng province had strong economic growth from 2009 to 2014 of 3.1% per annum, but growth has been low and slow from 2014 to 2019 at 1.1%. The COJ contributed the most to the provincial GDP of 44%, followed by the COT. The two peripheral regions of SDM and WRDM only contributed 4% and 3.8% to the province. The COT had the highest annual growth rate of 1.3% from 2014 to 2019. Secondly, Gauteng has a declining situation regarding GDP per capita with a negative growth rate of −1.2% from 2014 to 2019. Of the five municipal areas, both COJ and COT had much higher GDP per capita values and negative growth rates. SDM had the lowest GDP per capita at less than half of the two leading regions, namely COJ and COT. Thirdly, in terms of population density, Gauteng province had a density of 649 per sqkm and has an increased rate of 2.8% per annum. COJ and EKR metros had the highest densities of 2630 and 1562, with high levels of increases. The two more rural regions of SDM and WRDM have much lower at 208 and 176,


*Source: [37]. Note: () brackets contain annual growth percentages between observations.*

**111**

**Table 2.**

*An Assessment of the Impact of the Tourism Sector on Regional Economic Development…*

respectively. Fourthly, in terms of tourism jobs, Gauteng had 132 155 people working in tourism with a growth rate of 2.8% per annum. Compared with the five sub-regions, COJ contributes most with 39% of total jobs in the province, followed by EKR with 25% and COT at 24%. COT and EKR had the highest

Regarding disposable income, the Gauteng province has been increasing at 1.6% since 2014 in line with the slow growth of GDP growth. COJ again contributes most regarding the income of 38% of the total provincial income, followed by COT at 27%. SDM had the highest increase income of 2.1% coming from a low base, while the increase in income for the rest of the municipal regions had low growth rates of between 1.4–1.8%. Next, the international tourist trips are analysed as a ratio of total tourism trips to the regions. The WRDM had the highest ratio of 0.39, followed by EKR with a ratio of 0.34. All of the regions had relatively high growth rates of above 4.4% per annum. Tourism spending per capita for Gauteng province showed high growth levels from 2009 to 2014 at 7.4%, but as with the rest of the economy did the tourism sector also show much lower spending at 0.7% per annum since 2014 2019. COJ and COT had higher tourism spending per capita than the provincial values, with low growth rates of 0.21% and 0.4%, respectively. The three metro regions had much higher tourism spending levels if compared to the two more rural regions. By far, the WRDM had the highest increase in tourism spending per capita of 2.4% per annum since 2014. Lastly, the GDP per capita analysis indicates negative growth rates of – 2.5% per annum from 2014 to 2019, while only COJ has slightly higher tourism spending per capita than the province. As with the provincial growth rates, all of the municipal regions

**Table 2** indicates the correlation coefficients for the variables included in the econometric analysis. GDP per capita has a positive and significant relationship with all other variables, with GVA in the tourism sector with the highest coefficient of 0.93, followed by disposable income at 0.83. The two variables with the highest shared

> **Pop density**

**GVA in tourism sector**

**Jobs tourism sector**

**Int tourism trips**

**Spending in tourism per capita**

growth rates in jobs per annum of 4.4% and 3.6%, respectively.

also had negative growth rates of −3.18% to −4.17%.

**Disposable income**

0.9291 0.9414 0.7064 1.0000

0.7318 0.9506 0.8754 0.8630 1.0000

0.7943 0.9329 0.8362 0.9150 0.9733 1.0000

0.7017 0.8609 0.6326 0.7965 0.9020 0.9112 1.0000

**Variables GDP per** 

GDP per capita

Disposable income

GVA in tourism sector

Jobs in tourism sector

Int tourism trips

Spending in tourism per capita

*Correlation coefficients analysis.*

**capita**

1.0000

0.8314 1.0000

Pop density 0.5214 0.8374 1.0000

*DOI: http://dx.doi.org/10.5772/intechopen.95810*

#### **Table 1.**

*Descriptive analysis of municipal regions in Gauteng Province.*

#### *An Assessment of the Impact of the Tourism Sector on Regional Economic Development… DOI: http://dx.doi.org/10.5772/intechopen.95810*

respectively. Fourthly, in terms of tourism jobs, Gauteng had 132 155 people working in tourism with a growth rate of 2.8% per annum. Compared with the five sub-regions, COJ contributes most with 39% of total jobs in the province, followed by EKR with 25% and COT at 24%. COT and EKR had the highest growth rates in jobs per annum of 4.4% and 3.6%, respectively.

Regarding disposable income, the Gauteng province has been increasing at 1.6% since 2014 in line with the slow growth of GDP growth. COJ again contributes most regarding the income of 38% of the total provincial income, followed by COT at 27%. SDM had the highest increase income of 2.1% coming from a low base, while the increase in income for the rest of the municipal regions had low growth rates of between 1.4–1.8%. Next, the international tourist trips are analysed as a ratio of total tourism trips to the regions. The WRDM had the highest ratio of 0.39, followed by EKR with a ratio of 0.34. All of the regions had relatively high growth rates of above 4.4% per annum. Tourism spending per capita for Gauteng province showed high growth levels from 2009 to 2014 at 7.4%, but as with the rest of the economy did the tourism sector also show much lower spending at 0.7% per annum since 2014 2019. COJ and COT had higher tourism spending per capita than the provincial values, with low growth rates of 0.21% and 0.4%, respectively. The three metro regions had much higher tourism spending levels if compared to the two more rural regions. By far, the WRDM had the highest increase in tourism spending per capita of 2.4% per annum since 2014. Lastly, the GDP per capita analysis indicates negative growth rates of – 2.5% per annum from 2014 to 2019, while only COJ has slightly higher tourism spending per capita than the province. As with the provincial growth rates, all of the municipal regions also had negative growth rates of −3.18% to −4.17%.

**Table 2** indicates the correlation coefficients for the variables included in the econometric analysis. GDP per capita has a positive and significant relationship with all other variables, with GVA in the tourism sector with the highest coefficient of 0.93, followed by disposable income at 0.83. The two variables with the highest shared

