An Empirical Operationalization of Countries’ Destination Competitiveness Using Partial Least Squares Modeling

Abstract

Growth in tourism has resulted in escalating competition among destinations. Understanding destination competitiveness and its determinant factors is thus critical to tourism researchers and policy makers. Using partial least squares path modeling (PLSPM) on a cross-sectional sample of 154 countries, this study examines relationships among destination competitiveness and its predictors, including the economy, natural environment, and infrastructure. Results indicate that the economy has a positive, indirect impact on tourism competitiveness mediated through the infrastructure and the environment; moreover, infrastructure and environment have a direct, positive impact on tourism competitiveness. PLSPM was also used to compute composite scores for overall destination competitiveness, thus assigning rankings to the 154 countries assessed. This study contributes to extant theories on destination competitiveness, presenting important implications for policymakers on how to strengthen destination competitiveness, and providing an empirically based tool to help benchmark a country’s competitiveness in relation to other destinations.

Keywords

destination competitiveness partial least squares modeling (PLSPM)structural equation modeling (SEM)latent variables (LVs) scores cause indicators

Introduction

Tourism provides direct and indirect economic benefits to destinations through visitor spending, infrastructure development, and business investments (Fletcher and Archer 1991; Sinclair 1998). Many destinations, therefore, have invested heavily in developing their tourism industry in order to capture market share (Gooroochurn and Sugiyarto 2005). This has lead to a greater degree of competitiveness among destinations, particularly in regions that consider tourism as their major source of economic activity (Tsai, Song, and Wong 2009). Recent downturns in the global economy have created tougher market conditions, leading to greater levels of competition among tourism destinations (Assaf and Josiassen 2012). Understanding the factors that drive destination competitiveness is of particular interest to both researchers and policymakers (Tsai, Song, and Wong 2009). The Travel and Tourism Research Association (TTRA) recently developed a membership-wide agenda of priority topics in tourism research, identifying destination performance and competitiveness as one of the top two management research priorities critical to decision makers over the next decade (Williams, Stewart, and Larsen 2012).

Previous studies have presented several approaches to assessing the performance of tourism destinations and evaluating destination competitiveness and its determinants (e.g., Assaf and Josiassen 2012; Croes 2011). Some have used survey data of tourists’ visitor experiences to measure competitiveness (e.g., Crouch and Ritchie 1999; Dwyer and Kim 2003) and some have used secondary data to assess and compare competitiveness across countries (e.g., Gooroochurn and Sugiyarto 2005; Mazanec, Wober, and Zins 2007). In addition, others have used panel regression analysis to compare competitiveness of destinations in a specific geographical region, for example, small island nations in The Caribbean (Croes 2011).

Past studies have generally examined the relative importance of the determinants of competitiveness through exploratory statistical techniques; however, these approaches are limited in their ability to compute an actual score or index to represent and measure destination competitiveness or develop cross-country comparisons. There have been fewer attempts to examine the predictive relationships that may exist between the determinants of tourism performance, and how these factors influence the level of tourism competitiveness at the destination.

To fill these gaps, Assaker, Esposito Vinzi, and O’Connor (2011) identified several variables drawn from destination competitiveness studies into three underlying supply factors—the economy, infrastructure, and the environment. Assaker, Esposito Vinzi, and O’Connor (2011) identified the role that supply-side variables play in enhancing destination competitiveness, aligning their study with Crouch and Ritchie (1999), Dwyer and Kim (2003), and Dwyer et al. (2004). Assaker, Esposito Vinzi, and O’Connor (2011) used structural equation modeling (SEM) to examine a comprehensive causality model to understand tourism competitiveness. However, their study suffered from the same limitations as previous works in that it was unable to compute a score for each construct. Assaker, Esposito Vinzi, and O’Connor (2011) relied on SEM using maximum likelihood estimation (MLE) for their analysis; thus, they were only capable of testing “theory” rather than incorporating the predictive causes among the economy, infrastructure, environment, and tourism in computing the scores for each factor. Hence, an overall tourism competitiveness score could not be computed. Such a score would be particularly important for destination benchmarking.

This study attempts to address these limitations by proposing an alternative methodology, namely, partial least squares path modeling (PLSPM), to examine the network of predictive relationships among destination competitiveness and its determinant factors. PLSPM can examine how the latent variables or item measures are predicted by other variables and measures (Wold 1979). Thus, compared to traditional SEM techniques, for example, maximum likelihood estimation (MLE), PLSPM enables researchers to estimate the weight of indicators and regression coefficients across factors as well as estimate a score for each factor. This article will demonstrate how the use of PLSPM has numerous advantages over MLE-based SEM. For example, PLSPM necessitates fewer restrictive requirements than traditional SEM; can handle both reflective and formative specifications; has less stringent assumptions about the distribution of variables, rendering possible the analysis of structures when normality of data cannot be assumed; is resilient to problems of small sample size and can therefore be applied in situations when other methods are not viable. PLSPM is also considered to be the most adequate SEM technique to analyze complex models with higher numbers of latent and manifest variables in relation to the number of observations (Wold 1985). Such advantages are important in the specific context of our study.

In the case of destination competitiveness, PLSPM allows the computation of an aggregate score for competitiveness that can be used to rank destinations based on an aggregate competitiveness value. The authors are aware of the World Travel and Tourism Council’s (WTTC’s) initiatives to develop such an approach, for example, the WTTC Competitiveness Monitor (CM) 2004. Thus, this article will enable researchers and policymakers to understand what factors determine destination competitiveness, how indicators combine to form their respective factors, and how the factors correlate to determine destination competitiveness. Moreover, the study illustrates a method for computing an index for tourism competitiveness that can be used to benchmark destinations on a continual basis.

Destination Competitiveness Theory

Previous studies have presented various definitions of destination competitiveness (Crouch and Ritchie 1999, p. 137; D’Hauteserre 2000, p. 23; Hassan 2000, p. 239; Croes 2011). Heath (2003) posited that destination competitiveness “includes objectively measured variables such as visitor numbers, market share, tourist expenditure, employment, value added by the tourism industry, as well as subjectively measured variables, such as ‘richness of culture and heritage,’ ‘quality of the tourism experience, etc.” (p. 9). Furthermore, research on destination competitiveness has sought to develop a conceptual and theoretical basis to determine how destination competitiveness can be understood and measured (Crouch and Ritchie 1999; Dwyer and Kim 2003; Heath 2003) and apply models thus developed to data pertaining to destinations (Dwyer et al. 2004; Enright and Newton 2004, 2005; Gooroochurn and Sugiyarto 2005; Mazanec, Wober, and Zins 2007). The most influential models indicate that numerous determinants are relevant when examining destination competitiveness. The range of determinants these models have used are reviewed in the next section with the aim of selecting relevant constructs for the present study.

One of the earliest and most comprehensive frameworks of destination competitiveness comes from Crouch and Ritchie (Crouch and Ritchie 1999; Ritchie and Crouch 2000, 2003) and are based on Porter’s (1990) core diamond theory of competitive and comparative advantages. Crouch and Ritchie (1999) stipulated that competitiveness is determined by four types of components: (1) core resources and attractors including climate, cultural sites, special events, and tourism superstructure, which are the primary elements of destination appeal; (2) supporting factors and resources such as infrastructure and accessibility that enable a destination to build a successful tourism industry; (3) destination management, which includes destination management organizations’ activities to enhance the appeal of core andsupporting resources; and (4) qualifying and amplifying determinants, defined as situational conditions such as crimes, government fiscal policy, exchange rates, etc., which have the potential to negatively affect competitiveness. Crouch and Ritchie (1999) developed a comprehensive list of indicators that combine subjective consumer and objective industry measures for each component, resulting in the first initiative leading to a composite destination competitiveness index and a tool to simulate destination performance.

Dwyer and Kim (2003) introduced a holistic approach to the determinants and indicators of destination competitiveness. Their indicators were categorized into five subgroups: (1) endowed resources, such as mountains, scenery, and cuisine; (2) supporting factors, including infrastructure, service quality, and accessibility; (3) destination management, such as marketing strategy and safety regulations; (4) situational conditions such as crimes, industry structure, and world economic conditions; and (5) demand factors, including awareness, perception, and tourist preferences. Dwyer and Kim demonstrated how resources and destination management interact not only with situational conditions but also how tourism demand influences tourism destination competitiveness. Dwyer et al. (2004) then factor-analyzed 81 competitiveness items from Dwyer and Kim’s (2003) list to extract 12 principal components.

Previous studies have conceptualized destination competitiveness as a “house” composed of foundations, cement, building blocks, and a roof (Heath 2003). In a destination competitiveness framework, the “foundations” include key attractors such as personal safety and health, enablers such as infrastructure, value adders such as location and value for money, facilitators such as accommodation and airline capacity, and experience enhancers such as hospitality and authentic experiences. These provide the essential base for competitiveness. The “cement” includes stakeholders, communication, partnerships and alliances, information and research, and performance measurements that link the respective facets of competitiveness. The building blocks connect sustained destination competitiveness through an integrated development policy and strategic and destination marketing framework. Finally, the “roof” covers the human factor of destination competitiveness (Heath 2003).

Enright and Newton (2004, 2005) following Crouch and Ritchie’s (1999) conceptualization of destination competitiveness identified 37 business-related factors and 15 conventional factors to assess and measure the competitiveness of Hong Kong. Using survey data and factor analysis, they found strong empirical support for including both industry-level, or business-related, factors and destination attributes to explain competitiveness. This demonstrates the value of looking beyond tourism-specific factors and adopting a broader perspective to provide a holistic picture of destination competitiveness.

The above-mentioned studies have all relied to some extent on subjective qualitative data derived from surveys to evaluate competitiveness. Such studies cannot cover a wide range of countries because of the costs of collecting data and because certain intrinsic characteristics are common only to destinations that compete directly. Such characteristics differ among types of destinations, for example, the quality of beaches for rural destinations that have no access to the sea.

Based on a WTTC initiative, Crouch and Ritchie’s (1999) factor endowment, and the environment quality concept of Inskeep (1991) and Middleton (1997), Gooroochurn and Sugiyarto (2005) proposed 23 objective criteria that could be responsible for attracting tourists to destinations. These criteria came to be known as the Competitiveness Monitor (CM). The authors condensed the 23-item criteria into 8 summary indicators of price, economic and social impacts, human resources, infrastructure, environment, technology, openness, and social development. These summary indicators were used as an index to demonstrate the relative level of performance for each country. The researchers then used confirmatory factor analysis (CFA) with a sample of 93 countries to assign weights to each of the indicators on competitiveness. The analysis found that social and technology indicators had the highest weights whereas, surprisingly, human resources and environmental indicators had the lowest. Moreover, price had a significant inverse relationship with competitiveness. The authors argue that this list is far from exhaustive and the CFA implied that the destination competitiveness index appears to have a latent life of its own as a directly immeasurable feature of a destination, thereby causing or manifesting itself through the indicators.

Mazanec, Wober, and Zins (2007) criticized the validity of the CM, arguing that the competitiveness factor is a “formative” factor. They used the same eight CM indicators but applied formative schemes to calculate the various weights/coefficients between indicators. By estimating these weights, the authors paved the way for building a composite index for overall tourism competitiveness. However, both studies are limited to descriptive analysis and validation of the constructs rather than explaining and predicting causal relationships between constructs. Furthermore, the question as to how these relationships determine the level of a destination’s tourism competitiveness remains uncertain. Moreover, Mazanec, Wober, and Zins (2007) and Gooroochurn and Sugiyarto (2005) measured the underlying factors as a simple average of the indicators. Consequently, evidence of composite reliability, which demonstrates how strongly indicators explain their respective constructs, is lacking as fewer indicators were assigned to a multitude of factors, leading to a lower correlation among indicators.

In an attempt to advance the literature on destination competitiveness, Assaker, Esposito Vinzi, and O’Connor (2011) condensed 17 of the original 23 WTTC CM indicators into 4 factors: the economy, infrastructure, environment, and tourism demand. CFA was used to determine the weights of the indicators, which found that 15 of the 17 variables were significant in defining their assigned constructs. Two economic indicators, industry value added (IVA) and foreign direct investment (FDI), did not have significant loadings and were removed to achieve a reasonable model fit. Even after this deletion, economy indicators in the final model had loadings of less than 0.7, suggesting that causal relationships between economy items and the economy construct needed to be changed from reflective to formative; however, doing so led to identification problems under SEM.

Assaker, Esposito Vinzi, and O’Connor (2011) then used simultaneous equation models to test for the hypothetical causal relationships set between the factors. The economy construct (economy loadings < .7 even after purifying some indicators) had an indirect positive impact on tourism through the infrastructure and environment constructs. Infrastructure had a direct positive influence on environment. Moreover, the infrastructure and environment had a direct positive impact on generating tourism activities and revenues. Thus, the study verified causal relationships among constructs at the aggregate level using countries as the unit of analysis.

Finally, a recent study by Assaf and Josiassen (2012) attempted to assess the overall tourism competitiveness of various destinations. To do so, the authors identified 30 determinants, which they grouped into eight drivers of tourism performance (namely, tourism and related infrastructure; security, safety, and health; government policies; labor skills and training; economic conditions; tourism price levels; environmental sustainability; and natural and cultural resources). The authors used data from 129 countries and applied the data envelopment analysis (DEA) technique to compute the relative tourism performance of the 129 considered destinations. The data analysis identified the destination that represented the optimal or benchmark frontier.

The authors then ranked the remaining destinations based on how distant they were from the frontier (performing) destination. Next, the authors used truncated regression analysis to assess the impact of the 30 identified determinants of tourism performance. Despite the important contribution Assaf and Josiassen (2012) make to the literature by providing a tool to benchmark destinations based on a set of indicators, as well as assessing the importance of each, this study suffers from two major limitations. First, it does not account for causal relationships that may exist between the drivers/determinants regarding how they could impact each other. Second, several of the determinants were developed specifically within the context of their study and do not correspond to previously identified determinants. Furthermore, some of the determinants were not tested in the context of the existing destination competitiveness literature (e.g., WTTC CM; Crouch and Ritchie 1999; Gooroochurn and Sugiyarto 2005; Mazanec, Wober, and Zins 2007). As such, the determinants that Assaf and Josiassen (2012) developed need to be tested and validated further before they can be presupposed and used in other works on destination competitiveness.

Proposed Model and Variables

The model suggested in the present study is similar to that used in Assaker, Esposito Vinzi, and O’Connor (2011), which considers 23 indicators/determinants that were validated and tested in previous destination competitiveness studies (WTTC CM; Crouch and Ritchie 1999; Gooroochurn and Sugiyarto 2005; Mazanec, Wober, and Zins 2007). The proposed model examines the causal relationships among the economy, infrastructure, environment, and tourism; it, additionally, hypothesizes that the economy indicators form the underlying economy construct.

The proposed model is consistent with previous studies in economics, business and tourism that argue a formative structure for the economy construct (see Diamantopoulos and Winklhofer 2001; Ennew, Reed, and Binks 1993; Homburg, Workman, and Krohmer 1999; Johnston 1988). The prevalent practice of specifying a construct is based on classic theory, which assumes that an underlying latent construct causes observed variations in a construct’s measure. In such cases, the model is presented where the direction of the arrows goes from the construct to the indictors (Nunnally 1978). As such, the construct need not be explored further when reflective measures seem to behave well in reflecting their underlying construct. This is the case for the environment, infrastructure, and tourism competitiveness constructs as these were supported by earlier studies (Mazanec, Wober, and Zins 2007; Assaker, Esposito Vinzi, and O’Connor 2011). This was also confirmed in this study by the unidimensionality and internal consistency results of the exploratory factor and block analyses undertaken (discussed in the next section). Further analysis is required in certain cases where the traditional reflective scheme for a construct is considered conceptually inappropriate (Jarvis, Mackenzie, and Podsakoff 2003). This was the case for the economy construct.

Jarvis, Mackenzie, and Podsakoff (2003) developed four sets of criteria in determining whether a construct should be formative or reflective: (1) the direction of causality between the construct and its indicators; (2) the interchangeability of indicators; (3) covariation among indicators; and (4) whether or not indicators have the same antecedents and consequences, often referred to in SEM as the nomological net of the indicators. In following these criteria, it is apparent that the economy should be modeled as a formative construct. A destination’s economic situation is defined by its indicators, not vice versa, since a change in the indicators such as consumer price index (CPI), purchasing power parity (PPP), IVA, trade balance, and FDI affects the economy. Moreover, these factors do not necessarily correlate with each other (although they may actually covary in practice). For example, CPI has no reason to correlate with PPP, IVA, trade balance, or FDI. Furthermore, economy indicators do not necessarily share the same set of antecedents; for example, IVA is determined by factors that differ from factors affecting trade balance.

The measures used for each construct in the model were based on those used by Assaker, Esposito Vinzi, and O’Connor (2011). The latter changed the WTTC technology indicators, combining the Internet, phone and mobile device, and high-tech indexes. The high-tech index, which the WTTC system calculated based on the percentage of manufacturing product exports characterized by high R&D intensity, is not a clear driver for a destination to become more successful in tourism. Mazanec, Wober, and Zins (2007) applied a similar solution. The infrastructure indicator, which is defined as the share of population with access to improved sanitation and drinking water, was aggregated as such data were communicated in aggregate form in most country statistics. The human resource indicator and social indicators with the exception of income per capita index of the social indicators were combined with technology and infrastructure to create the infrastructure construct, as the latter is measured by the level of technological advancement and access to both transportation and utilities infrastructure. Income per capita was deselected from social indicators as it was found to be highly correlated with all other indicators, so deleting it reduced multicollinearity across latent factors (Mazanec, Wober, and Zins 2007).

The WTTC’s price and openness indicators were grouped to form the economy factor. The three indices of hotel prices, visa openness, and taxes on international trade were removed and replaced by FDI, IVA, and CPI. Hotel price was removed because CPI arguably is a better proxy for tourism costs (Mangion, Durbarry, and Sinclair 2005; Song and Witt 2003). Using CPI is justified by the complex nature of the tourism coproduct, as tourists consume both enclave and traditional commodities. Visa openness was also removed as it was unrepresentative of any of the underlying tourism supply variables examined. Finally, taxes on the international trade index were removed because the openness indicators already included level of international trade.

However, in contrast to Assaker, Esposito Vinzi, and O’Connor (2011), the economy factor was assumed to follow a formative scheme. The environment construct was represented by CO₂, treaties, and electricity production. The remaining variables were retained as the literature suggested, although the human tourism indicator comprising the number of tourist arrivals and expenditures was relabeled “the tourism competitiveness” factor to investigate an explanatory system of tourism destination competitiveness.

This study builds on Mazanec, Wober, and Zins’s (2007) work where they argue the need to explain and measure the model’s final outcome, that is, destination competitiveness. As such, this study uses the total number of tourist arrivals and tourist expenditures/receipts to measure the competitiveness construct in the context of a comprehensive system of cause–effect relationships. The performance/competitiveness of tourism destination management can be measured using a variety of indicators, such as the extent to which natural and cultural resources are preserved (e.g., Inskeep 1987; Ritchie et al. 2001) or visitors’ satisfaction with the tourism product provided (e.g., Kozak 2002). These measures need to be collected using survey data of tourists’ and destination managers’ perceptions of their experiences at different destinations. This makes it quasi-impossible to collect data across a large range of countries. Therefore, tourist arrivals and expenditures were used in the present study as measures of tourism performance and to operationalize the destination competitiveness construct (see Mazanec, Wober, and Zins 2007; Assaker, Esposito Vinzi, and O’Connor 2011). The seventeen variables analyzed in this model are explained in Table A1 in the appendix.

Model Analysis with PLS Path Modeling

Our conceptual analysis involves rearranging the WTTC competitiveness indicators into a predictive model by relating the WTTC indicators to a set of dependent latent variables—namely, economy, infrastructure, environment, and tourism demand—to extend the model into simple theory. Rearranging the WTTC indicators presents a comprehensive framework for predicting tourism demand at the destination. The structural model presented in this study is examined using PLSPM (Wold 1979). This is a partial information method and can maximize the explained variance of all dependent variables, making it suitable for prediction-oriented goals in situations of high complexity but low theoretical information (Jöreskog and Wold 1982). PLSPM is achieved through a two-step process: (1) an iterative estimation of latent variables scores by estimating outer and inner weight using the PLS algorithm; (2) estimation of loadings and path coefficients through OLS regression (Tenenhaus 1998). PLSPM estimates the latent variables as exact linear combinations of the observed measures, thus, overcoming indeterminacy problems and providing an exact definition of component scores. PLSPM is used for our analysis since (1) the data used were found to be nonnormally distributed, (2) the sample size of 154 countries is small compared to the 17 indicators used, (3) the model contains formative and reflective indicators, and (4) the theoretical underpinnings of the relationships posited are uncertain.

Hypotheses

The variables comprising the constructs in this study play an important role in detecting the overall relationships for which the model accounts. For example, economic factors can have a positive or negative impact on tourism arrivals (Ioannides and Debbage 1998). Both the CPI and PPP are predicted to have a direct negative impact on the tourism competitiveness construct since traditional economic theory suggests that absolute and relative prices at a destination will determine the level of tourism to that destination (Ioannides and Debbage 1998). As such, if prices at a destination increase, international tourism becomes more expensive, resulting in a decreased travel inflow into that country, and vice versa (Dwyer and Kim 2003; Song and Witt 2003). However, trade volume, FDI, and IVA are expected to have an indirect positive effect on tourism arrivals. When FDI is channeled into a territory with value added from manufacturing, the tourism infrastructure of a country is developed, thereby indirectly influencing tourism arrivals. As a result, FDI and IVA are assumed to have a direct positive impact on infrastructure, consequently affecting tourism (Boniface and Cooper 1987). Trade volume and FDI impact the well-being of the host community by providing capital, technology transfer, superior wages, and improved infrastructure conditions, thereby affecting tourism. As a consequence, these are affected by economic and infrastructural results (Ioannides and Debbage 1998; Quazi 2007). This forms the basis for the following hypotheses:

Hypothesis 1: A significant negative relationship exists between economy and tourism competitiveness.

Hypothesis 2: A significant positive relationship exists between economy and infrastructure.

It is expected that there is a positive relationship between infrastructure and tourism. The more developed a given country’s infrastructure, the greater the possibility that the country may be able to tap into its tourism potential (Crouch and Ritchie 1999; Khadaroo and Seetanah 2007). However, high investment in infrastructure may negatively impact the environment, which negatively affects tourism spending. Since the environment construct contains negative indicators of the physical and natural states such as carbon emissions and electricity production, a positive relationship is assumed between the infrastructure and the environment construct. In addition, a negative relationship is assumed between the environment and tourism:

Hypothesis 3: A significant positive relationship exists between infrastructure and tourism competitiveness.

Hypothesis 4: A significant positive relationship exists between infrastructure and environment.

Hypothesis 5: A significant negative relationship exists between the environment and tourism competitiveness.

The theoretical model with all the factors, underlying variables, and hypothesized predictive relationships is illustrated in Figure 1.

Figure 1.

Proposed theoretical models with formative economy construct.

Research Methodology

Research Design and Data

An initial sample of 204 countries was selected and obtained from Euromonitor International in 2009. Countries with more than 20% of missing entries on the 17 variables were deleted from the analysis (Hair et al. 2010) resulting in 154 countries. The nearest-neighbor approach was used (Olinsky, Chenb, and Harlow 2003) to impute missing entries and arrive at the final data set. Moreover, given that the measures have different scales and that the standard deviations of the selected variables reflected large values, the data were standardized by subtracting the mean and dividing by the standard deviation so as to remove the scale effect of the variables prior to investigating the relationships. Such standardization provides a more accurate approach to comparing countries across the same variable and more adequately compares the predictive relationships of variables in the model (Russellet al. 1998). Moreover, the large values of standard deviation within the variables indicate that the data were not normally distributed as countries ranged from highly developed to less developed.

Analysis of Results

Exploratory factor analysis (EFA) was conducted on the unstandardized data set to support the theoretical verification of our constructs (Hair et al. 2000). Following the EFA, we tested the unidimensionality of each construct by conducting a block factor analysis and reliability analysis for each construct separately to verify whether each construct was sufficient to influence the set of responses/indicators identified from previous literature and proposed in this study. Once the unidimensionality and internal consistency of each factor were verified, we could then perform PLSPM analysis following two steps: (1) validating the outer model primarily through convergent and discriminant validity as well as content validity for reflective and formative schemes and (2) fitting the inner model through latent variable path analysis (Chin 1998).

Research Results

Exploratory Factor Analysis

Principal components analysis (PCA) was conducted on the entire unstandardized data set by running an oblique (i.e., Promax) rotated analysis. Promax rotation maximizes the variance on the new axes while allowing factors to be correlated with one another (Fabrigar et al. 1999). This outcome is desirable as we want to identify predictive relationships between constructs; as such, they need to be correlated. The final model structure from EFA analysis was found to explain 76.58% of the variance (see Table 1).

Table 1.

Total Variance Explained.

	Eigenvalues from Promax Rotation
Component	Total	% Variance	Cumulative %
1	7.171	42.180	42.180
2	2.226	13.096	55.276
3	1.351	7.949	63.225
4	1.219	7.173	70.398
5	1.052	6.188	76.586

Note: Extraction method: principal components analysis; rotation method: Promax with Kaiser normalization.

Table 1 reports the eigenvalues after the rotation where total variance was found to be split among five constructs with eigenvalues >1. Further investigating the loadings (related to the 17 variables) on the five factors showed that the only identified difference was the economy construct, which was found to be represented by two factors rather than one suggesting that economic variables could represent alternative aspects of the economy construct. The five extracted factors and their corresponding indicators/variables are:

Infrastructure—seven items: television sets index (TV), computers index (PC), internet hosts index (INT), road index (ROAD), number of vehicles index (AUTOS), newspaper index (NEWS), and sanitation facilities index (SAN). It accounts for 42.18% of the total variance extracted.

Environment—three items: electricity production index (ELEC), CO₂ emissions index (CO₂), and signed environmental treaties (TRE). It accounts for 13.09% of the total variance extracted.

Destination competitiveness—two items: tourists’ arrivals (TA) and tourists’ expenditures (TE). It accounts for 7.94% of the total variance extracted.

First economy construct—three items: consumer price index (CPI), purchasing power parity (PPP), and industry value added (IVA). It explains 7.17% of the total variance extracted.

Second economy construct—two items: foreign direct investment (FDI) and trade (TRA). It explains 6.18% of the variance extracted.

Exploratory Block Factor and Reliability Analysis

We tested the dimensionality of each construct by conducting a PCA of the standardized data of the four blocks of variables. Tourism, infrastructure, and environment constructs were unidimensional, each represented by one factor with an eigenvalue >1. For the economy, several factors had eigenvalues close to 1 suggesting that the items could represent alternative aspects of the construct. In addition, all factor loadings with the exception of the economy performed well inside each block, with loadings >0.7 further supporting the unidimensionality of the blocks. All factors inside each block fell within a relatively small range: 0.68 to 0.91 for infrastructure, 0.81 to 0.96 for environment, and 0.91 for tourism. For the economy, the same five indicators loaded on three different dimensions with CPI and IVA on one factor, FDI and trade balance on another, and PPP on a third factor (Table 2). This suggests that economy is a multidimensional construct, with CPI, IVA, and PPP representing different dimensions of the economy than those represented by FDI and Trade Balance. These latter factors most likely represent a country’s degree of economic openness (Ioannides and Debbage 1998).

Table 2.

Factor Matrix, Cronbach’s α, Composite Reliability, and Eigenvalues by Variable Blocks Using Component Analysis Extraction Method.

Constructs	Variables	Factor 1	Factor 2	Factor 3	Critical value	Eigenvalues	Cronbach’s α	Dillon-Goldstein’s ρ (CR)
Economy^a,b	PPP			0.98	1	1.31
	CPI	0.77				1.26
	IVA	0.78				0.96
	FDI		0.79			0.77
	TRADE		0.79			0.72
Infrastructure	PC	0.88			1	4.92	0.92	0.94
	TV	0.88				0.86
	INT	0.91				0.52
	ROAD	0.76				0.42
	AUTOS	0.90				0.27
	SAN	0.68				0.15
	NEWS	0.78				0.09
Environment	ELEC	0.95			1	2.45	0.89	0.93
	CO2	0.96				0.47
	TRE	0.81				0.06
Tourism	TA	0.91			1	1.65	0.80	0.91
	TEXP	0.91				0.33

Principal components analysis with rotation method: varimax with Kaiser normalization.

Absolute loading values less than .25 are not shown.

These findings further support our representation of the economy as a formative construct where changes in the indicators are expected to cause changes in the construct itself (Jarvis, Mackenzie, and Podsakoff 2003). Empirical support also indicates that infrastructure, the environment, and tourism competitiveness should be represented as reflective constructs (Nunnally 1978). Finally, the Cronbach’s alpha and Dillon-Goldstein’s rho for infrastructure, environment, and tourism competitiveness constructs were robust and well above the lower limit of 0.7 (Nunnally and Bernstein 1994), indicating high-scale reliability and supporting the unidimensionality and reflective scheme of these factors (Table 2).

Partial Least Squares Analysis

PLSPM was conducted using XLSTAT-PLSPM (Addinsoft 2011) on the full data set of the standardized data, using Mode A, the reflective scheme, for infrastructure, environment, and tourism competitiveness and Mode B for the formative economy construct. The centroid scheme is also used for the estimation of inner weights.

Outer model analysis

First, the formative and reflective measurement models were analyzed. PLSPM makes no distributional assumptions; thus, only non-parametric tests can be used to evaluate the explanatory model (Chin 1998). In the proposed model (Figure 1), three reflective constructs were examined: infrastructure, environment, and tourism competitiveness. The usual tests were applied and the convergent validity of the constructs was supported since all factor loadings exceed the threshold level of 0.7 (Table 3); thus, more than 50% of the variance in the observed variable was due to the underlying construct (Hulland 1999). Furthermore, the bootstrap test showed high significance levels for all loadings at the .05 bootstrapping significance level (see Table 4.). The average variance extracted (AVE)—which measures the amount of variance of the indicator accounted for by the construct relative to the amount due to the measurement—achieved values of 0.674, 0.821, and 0.832 for infrastructure, environment, and tourism, respectively. With respect to discriminant validity, the average shared variance of a construct and its indicators should exceed the shared variance with every other construct of the model. Therefore, the root of AVE should surpass the correlation coefficient of the construct with every other construct of the model. This is the case in our outlined model (Table 5).

Table 3.

Results of the Outer Model: Latent Variables and Formative Economy Indicators.

Latent Variable	Manifest Variables	Standardized Loadings	Standardized Loadings (Bootstrap)	Critical Ratio (CR)	Lower Bound (95%)	Upper Bound (95%)	Outer Weight	Outer Weight (Bootstrap)	Critical Ratio (CR)	Lower Bound (95%)	Upper Bound (95%)
Economy	PPP	−0.479	−0.485	−4.317	−0.754	−0.283	−0.546	−0.583	−3.081	−0.979	−0.314
	CPI	0.03	−0.012	0.202	−0.362	0.319	−0.144	−0.249	−0.453	−1.248	0.166
	IVA	0.776	0.715	6.918	0.439	0.873	0.820	0.772	8.708	0.594	0.934
	FDI/GDP	−0.282	−0.283	−2.124	−0.582	−0.04	−0.286	−0.297	−2.112	−0.666	−0.083
	TRA/GDP	0.137	0.058	0.509	−0.439	0.627	0.187	0.100	0.714	−0.560	0.594
Infrastructure	PC	0.834	0.83	15.758	0.696	0.914	0.140	0.140	7.072	0.100	0.183
	TV	0.855	0.851	23.163	0.774	0.91	0.162	0.163	6.802	0.127	0.226
	INT	0.863	0.856	20.653	0.744	0.921	0.137	0.138	5.912	0.095	0.203
	ROAD	0.775	0.777	16.778	0.676	0.869	0.186	0.185	8.582	0.134	0.236
	AUTOS	0.853	0.85	17.353	0.729	0.93	0.141	0.142	6.057	0.101	0.202
	SAN	0.73	0.735	13.101	0.587	0.819	0.203	0.204	5.054	0.136	0.299
	NEWS	0.83	0.831	22.866	0.741	0.889	0.229	0.229	5.527	0.158	0.320
Tourism	T.Arrivals	0.904	0.922	17.015	0.849	0.998	0.525	0.604	1.900	0.244	1.411
	T.EXP	0.92	0.917	18.064	0.769	0.986	0.571	0.533	5.310	0.324	0.854
Environment	CO₂	0.941	0.913	10.488	0.623	0.983	0.379	0.360	10.282	0.281	0.438
	Elec	0.95	0.929	17.133	0.789	0.985	0.369	0.350	7.633	0.258	0.493
	TRE	0.821	0.85	12.457	0.765	0.991	0.357	0.433	1.555	0.175	1.094

Note: Values of their outer weights (iteratively calculated via the partial least squares algorithm) are used to calculate scores of latent variables as noted on the arrows connecting manifest variables to their composing blocks in Figure 2, as discussed later in the analysis section.

Table 4.

Results of Discriminant Validity: Latent Variables and Formative Economy Indicators (Squared Correlations for any Pair of Latent Variables < AVE).

	Economy	Infrastructure	Environment	Tourism	Mean Communalities (AVE)
Economy	1	0.120	0.042	0.032	0.186
Infrastructure	0.120	1	0.381	0.505	0.674
Environment	0.042	0.381	1	0.711	0.821
Tourism	0.032	0.505	0.711	1	0.832
Mean communalities (AVE)	0.186	0.674	0.821	0.832	0

Note: AVE = average variance extracted.

Table 5.

Results of the Inner Model.

	Path Coefficient	Standard Error	t	Pr > \|t\|
Economy → Infrastructure	0.347	0.076	4.555	0.000
Infrastructure → Environment	0.617	0.064	9.676	0.000
Infrastructure → Tourism	0.308	0.050	6.193	0.000
Environment → Tourism	0.653	0.050	13.144	0.000
Economy → Tourism	−0.067	0.046	−0.937	0.350

For the economy construct, content validity was examined at both indicator and construct levels. At the indicator level, the results of the bootstrap tests showed high significance levels for PPP, IVA, and FDI loadings on the economy construct at the .05 bootstrapping significance level (Table 3). The variance inflation factor (VIF) for the economy indicators showed levels lower than 2 (1.011, 1.073, and 1.080 for PPP, IVA, and FDI, respectively), suggesting that these three indicators are not highly correlated to one another. Therefore, PPP, IVA, and FDI indicators were retained in the outer model measurement model. The achieved explained variance, or R², of the endogenous constructs was primarily used to determine whether a theoretically sound exogenous factor was appropriately operationalized (Diamantopoulos and Winklhofer 2001). This is further discussed in the next section.

Inner model analysis

The R² results of the tested model demonstrated that a substantial part of the variance of the latent constructs can be explained by the model. Thus, the CPI, PPP, and Trade Balance contribute to a fair extent to forming the economy construct. The cross-sectional regressions for infrastructure, environment, and tourism (0.16, 0.38, and 0.89) achieved an explained variance of 15% to 30%, thus supporting the nomological validity of the model (Chin 1998).

Another assessment of the structural model involves its ability to predict the endogenous latent variable indicators, referred to as the cross-validated redundancy measures (Wold 1982). To this end, the Stone-Geisser Q2 values (Stone 1974; Geisser 1975), which is the predominant measure of predictive relevance measured using blindfolding procedures (Tenenhaus et al. 2005), were examined. Q2 values for the infrastructure and environment variable indicators were larger than zero, suggesting predictive relevance in explaining the endogenous latent variables under evaluation. Furthermore, Q2 values for the tourism indicators were all above 0.35, indicating substantial predictive relevance in explaining the tourism variables (Henseler, Ringle, and Sinkovics 2009).

Finally, a bootstrap with N = 1,000 samples was run, providing t values and two-tailed significance levels for the estimates of the path coefficients (Davison and Hinkley 1998). Table 5 illustrates the results of the inner model with the results of the conducted bootstrap, showing that three of the five hypotheses were supported. The infrastructure construct was positively influenced by the state of the economy (regression coefficient = .347), supporting hypothesis 2. Environment was positively influenced by infrastructure (regression coefficient = .617), supporting hypothesis 3. Tourism was positively influenced by infrastructure (regression coefficient = .308), supporting hypothesis 4. Finally, tourism was positively influenced by environment (regression coefficient = .653), leaving hypothesis 5 unsupported. The positive relationship between environment and tourism is somewhat surprising given the measurements used. In addition, the economy construct was found to have no influence on tourism, demonstrating the nonsignificance of hypothesis 1. Based on these results, we can conclude that the initially hypothesized model depicted in Figure 1 is a good fit for the data; this is despite the nonsignificance of the loadings of some of the economy variables and the path coefficient between the economy and tourism. The model provides sound predictive ability for the estimated endogenous latent variables and their underlying indicators as well as a robust methodological approach to computing latent variable scores.

Estimating the Aggregate Tourism Competitiveness Score

The inner model represented in Figure 2 shows how endogenous latent variables are connected to other latent variables in the model. In particular, the values appearing on the arrows connecting the latent variables correspond to the regression coefficients of the scores in the regression of the endogenous on exogenous factors. Scores of latent variables are defined in Figure 2 based on the values of their outer-weights, which are iteratively calculated using the PLS algorithm. Given this study’s rationale, the following equation allowed us to calculate the aggregate scores for destination competitiveness for each country directly from the original data table using the tourism competitiveness indicator measurements and the outer weights of the tourism indicators on which the tourism competitiveness factor is regressed.

Figure 2.

Results of the inner model regression coefficients as well as outer model weights of a tourism competitiveness framework with a formative economy construct.

Destination aggregate competitiveness score = . 525 \times TA + . 571 \times TEXP

The countries with the 10 highest and 10 lowest aggregate scores for destination competitiveness are shown in Figure 3.

Figure 3.

Ranking of top and base countries based on Destination Competitiveness scores predicted using the partial least squares path modeling (PLSPM) structural model.

Discussion and Conclusions

This research examined a network of predictive relationships among destination competitiveness and its determinant factors. The analysis was based on a sample of 154 countries. It found that economic factors, environment, and infrastructures play an influential role in determining destination competitiveness. PLSPM analysis was used to calculate destination competitiveness scores for the 154 countries. France, the United States, and China achieved the highest rating, whereas Macedonia, Madagascar, and Gambia achieved the lowest ranking.

The predictive model examined in this study presented the economy construct as a formative rather than reflective construct. This study demonstrated that infrastructure and environment have a direct, positive impact on generating tourism activities and revenues, with the environment construct contradicting the hypothesized effect of a negatively defined environment paradigm on tourism. Countries that scored high in "competitiveness" are actually industrialized countries with high organizational structures. These findings support those of Assaf and Josiassen (2012), who found that high-performance countries have well-established tourism industries.

However, these countries also produce higher CO₂ emissions and have a higher degree of electricity production. This could explain the positive relationships found between environment and competitiveness. These results partially contradict Gooroochurn and Sugiyarto’s (2005) findings that society and technology indicators have the highest weight in defining destination competitiveness whereas environment indicators have the lowest weight. The relationship between the economy and tourism was nonsignificant and in contrast to our expectations. The most plausible explanation for this finding is the composition of the economy construct that is used. The present study found that IVA and FDI were the significant variables that comprise the economy construct (with CPI being nonsignificant). Given that FDI channeled into an economy and value added from manufacturing are assumed to have an indirect positive impact on tourism through infrastructure development rather than a direct effect on tourism, this could explain the nonsignificance of the relationship between the economy and tourism, which is driven mainly by price levels at destinations (measured through the CPI variable, which was found to be nonsignificant in this case).

This research presents valuable implications for policymakers and destination managers as it illustrates the causal relationships between the factors driving tourism and destination competitiveness. It also identifies the factors comprising the economy, infrastructure, and the environment, which in turn impact upon tourism arrivals and revenues.

This provides practical applications for policymakers as it helps them make better decisions on the allocation of resources by understanding and measuring the impacts of the identified variables. For example, the results from the analysis demonstrate that policymakers should pay particular attention to FDI, IVA, and PPP levels, not only in terms of how these indicators contribute to economic growth, but also how developing the infrastructure and tourism activities and revenues are affected by economic outcomes. High levels of PPP and value added from manufacturing create a positive economic environment and lead to positive social improvements and developmental progress. Consequently, these have a positive effect on a country’s tourism competitiveness.

FDI, however, was found to have a negative impact on social improvement and developmental progress because of the difficulties in directing foreign investment in areas that may be most beneficial to the economy and locals. This will ultimately present a negative impact on tourism competitiveness. The study demonstrated the need for countries to enhance their infrastructure factors in order to increase tourism arrivals and revenues. The extent to which the Internet and other wireless communication technologies parallel tourism development was confirmed, with internet considered to be an important driver of tourism and facilitating the access of tourists to the destination.

Finally, PLSPM enabled us to calculate an aggregate tourism competitiveness index across countries. This enables destinations to benchmark their tourism competitiveness relative to other countries. They can do so by adjusting the outer weights of the tourism indicators used to compute destination competitiveness based on how the structural model is specified for a country’s underlying economic, infrastructural, and environmental factors.

This is possible due to an inherent property of PLSPM that the outside weights used to compute the tourism competitiveness latent factor, which in this case are the weights of tourist arrivals and expenditures, are derived from an iterative process, which also considers the internal weights (relationships) between the constructs (e.g., economy, infrastructure, environment, and tourism). In this study, both the inner (causes of destination competitiveness) and outer weights (consequences of destination competitiveness) were used to compute a destination’s competitiveness.

The proposed model and approach thus incorporates the predictive relationships among the economy, infrastructure, environment, and tourism in computing an overall tourism competitiveness score for a destination. This is superior to simply looking at the number of tourists or tourists’ receipts to rank destinations, as is commonly used by the World Tourism Organization (UNWTO).

Limitations and Future Research Direction

The tourism competitiveness score presented in this study assumes that the effects of tourist attractions are constant across countries. Thus, this ranking simply tells us about tourism competitiveness or the potential of each country based on its underlying economic, infrastructural and environmental factors, disregarding how much each country taps into its superstructure, or the specific hospitality and tourism features it has created to render the destination more attractive. Thus, based on the scores, if the United States, for example, gives complete importance to its tourism sector, it has the highest potential and currently occupies the best position in terms of tourism competitiveness. This constraint provides an opportunity for future research. The model can be improved by including additional variables such as the number of hotels, restaurants, and attractions in each country. These additional variables may help strengthen the model and provide another trajectory for future research.

The underlying dimensions of the latent factors used in this study are not exhaustive and were derived from the limited number of previous studies available. For example, little is known about environmental quality and its indictors. This may explain the unexpected positive relationship between the environment and tourism. Future studies could utilize subjective judgments to measure environmental quality, perhaps collecting data via surveys.

Future research could also comprehensively examine the economy measures as this would help verify the complexity of the relationship between the economy indicators and those underlying the infrastructure, environment, and tourism. Such research would address issues identified in this study in terms of the nonsignificant relationship between the economy and tourism. For example, using measures other than CPI, or using different ways to measure CPI, such as moving average, could extend this research and potentially improve model validity (see, Dwyer, Forsyth, and Rao 2000). Subjective measures of travelers’ perception of price levels at a destination could also be used to further enhance the composition of the economy construct (Dwyer and Kim 2003). Despite these limitations, this study provides a valuable contribution to the tourism literature and expands on existing theories on tourism destination competitiveness and its predictors. It presents a tourism competitiveness index and assigned rankings to 154 countries worldwide.

Footnotes

Appendix

Table A2.

Ranking of Countries Based on their Aggregate Destination Competitiveness Score Estimated Using Results of the Partial Least Squares Path Modeling (PLSPM) Structural Model.

Rank	Country	Destination Competitiveness Score
1	United States	8.650
2	France	4.214
3	China	3.681
4	Spain	3.417
5	Italy	2.743
6	Germany	1.768
7	United Kingdom	1.445
8	Mexico	1.414
9	Turkey	1.157
10	Japan	1.112
11	Austria	0.914
12	Malaysia	0.879
13	Russia	0.871
14	Canada	0.690
15	Thailand	0.483
16	Hong Kong, China	0.482
17	Greece	0.460
18	Sweden	0.452
19	Singapore	0.395
20	Netherlands	0.370
21	Australia	0.368
22	Switzerland	0.326
23	Portugal	0.282
24	Tonga	0.252
25	India	0.249
26	Egypt	0.240
27	Ukraine	0.228
28	Poland	0.220
29	Macau	0.213
30	Saudi Arabia	0.158
31	South Korea	0.111
32	Denmark	0.057
33	Morocco	0.055
34	South Africa	0.051
35	Croatia	0.041
36	Hungary	0.033
37	Brazil	0.029
38	Belgium	0.013
39	Kyrgyzstan	−0.012
40	United Arab Emirates	−0.012
41	Senegal	−0.068
42	Romania	−0.068
43	Ireland	−0.072
44	Norway	−0.074
45	Czech Republic	−0.082
46	Indonesia	−0.088
47	Tunisia	−0.098
48	Syria	−0.128
49	Argentina	−0.130
50	Bulgaria	−0.140
51	Tuvalu	−0.140
52	Finland	−0.157
53	Kazakhstan	−0.157
54	Taiwan	−0.175
55	Dominica	−0.212
56	Vietnam	−0.219
57	Dominican Republic	−0.223
58	Jordan	−0.242
59	Luxembourg	−0.242
60	Malta	−0.242
61	Iceland	−0.243
62	New Zealand	−0.250
63	Philippines	−0.253
64	Belarus	−0.254
65	Chile	−0.281
66	Israel	−0.283
67	Peru	−0.297
68	Lebanon	−0.301
69	Cuba	−0.317
70	Colombia	−0.319
71	Cyprus	−0.329
72	Mozambique	−0.334
73	Iran	−0.337
74	Uruguay	−0.337
75	Costa Rica	−0.339
76	Vanuatu	−0.339
77	Cambodia	−0.342
78	Jamaica	−0.346
79	Slovenia	−0.347
80	Paraguay	−0.347
81	Djibouti	−0.348
82	Estonia	−0.348
83	Guyana	−0.348
84	Mongolia	−0.348
85	New Guinea	−0.348
86	Suriname	−0.348
87	Guatemala	−0.348
88	Zimbabwe	−0.350
89	Azerbaijan	−0.358
90	Slovakia	−0.364
91	Albania	−0.365
92	Armenia	−0.365
93	Georgia	−0.365
94	Qatar	−0.370
95	Bahamas	−0.370
96	Botswana	−0.373
97	Pakistan	−0.373
98	Lithuania	−0.377
99	Panama	−0.381
100	Algeria	−0.383
101	Kuwait	−0.383
102	Oman	−0.384
103	Latvia	−0.384
104	Moldova	−0.384
105	Bangladesh	−0.386
106	Congo-Brazzaville	−0.386
107	Guinea	−0.386
108	Mauritania	−0.386
109	Niger	−0.386
110	Nigeria	−0.386
111	Laos	−0.387
112	Zambia	−0.388
113	Chad	−0.389
114	Kenya	−0.389
115	Togo	−0.389
116	Venezuela	−0.389
117	Ecuador	−0.393
118	Malawi	−0.394
119	Ethiopia	−0.395
120	Swaziland	−0.395
121	El Salvador	−0.396
122	Mauritius	−0.400
123	Namibia	−0.402
124	Nicaragua	−0.404
125	Uganda	−0.404
126	Honduras	−0.404
127	Tanzania	−0.404
128	Ghana	−0.404
129	Serbia	−0.413
130	Fiji	−0.414
131	Uzbekistan	−0.414
132	Maldives	−0.415
133	Bolivia	−0.416
134	Rwanda	−0.417
135	Barbados	−0.417
136	Seychelles	−0.418
137	Nepal	−0.424
138	Yemen	−0.427
139	Sri Lanka	−0.427
140	Sudan	−0.429
141	Benin	−0.434
142	Burkina Faso	−0.434
143	Burundi	−0.434
144	Cameroon	−0.434
145	Mali	−0.434
146	Cape Verde	−0.436
147	Bosnia-Herzegovina	−0.436
148	Libya	−0.436
149	Trinidad and Tobago	−0.436
150	Bahrain	−0.437
151	Bhutan	−0.437
152	Macedonia	−0.437
153	Madagascar	−0.442
154	Gambia	−0.444

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) received no financial support for the research, authorship, and/or publication of this article.

Author Biographies

Guy Assaker, PhD, is an Assistant Professor at the School of Business, Lebanese American University. He concurrently holds an adjunct research fellowship position at the Center for Tourism and Leisure Management at the University of South Australia. His research focuses on the use of advanced multivariate statistical methods to model hospitality and tourism subjects.

Rob Hallak, PhD, is a Lecturer in Management at the UniSA Business School, University of South Australia. His research focuses on entrepreneurship and the performance of small and medium enterprises in the tourism and hospitality sector.

Vincenzo Esposito Vinzi, PhD, is Professor of Statistics at ESSEC Business School, France. His research includes multivariate statistics, factorial methods, structural equation modeling, partial least squares regression, and path modeling, with business and industry-oriented applications.

Peter O’Connor, PhD, is Professor of Information Systems at ESSEC Business School France, where he also serves as Academic Director of ESSEC’s MBA Hospitality Management programs. His research focuses on technology, distribution, e-commerce, and electronic marketing applied to hospitality and tourism.

References

Assaf

A. G.

Josiassen

(2012). “Identifying and Ranking the Determinants of Tourism Performance: A Global Investigation.” Journal of Travel Research, 51 (4): 388-99.

Assaker

Esposito Vinzi

O’Connor

(2011). “Modelling a Causality Network for Tourism Development: An Empirical Analysis.” Journal of Modelling in Management, 6 (3): 258-78.

Boniface

Cooper

(1987). Worldwide Destinations: The Geography of Travel and Tourism. Boston, MA: Butterworth-Heinemann.

Buhalis

, (1998). “Strategic use of information technologies in the tourism industry.” Tourism Management, 19(5): 409-421.

Buhalis

, (2000). “Marketing the competitive destination of the future.” Tourism Management, 21(1):97-116.

Chathoth

(2001). “Causal Modelling Methodology in Tourism: An Empirical Analysis.” Journal of Services Research, 1 (2): 49-72.

Chin

W. W.

(1998). “The Partial Least Squares Approach for Structural Equation Modeling.” In Modern Methods for Business Research, edited by Marcoulides

G. A.

London: Lawrence Erlbaum, pp. 236-95.

Croes

(2011). “Measuring and Explaining Competitiveness in the Context of Small Island Destinations.” Journal of Travel Research, 50 (4): 431-42.

Crouch

Richie

J. R.

(1999). “Tourism, Competitiveness, and Societal Prosperity.” Journal of Business Research, 44: 137-52.

10.

Davison

A. C.

Hinkley

D. V.

(1998). Bootstrap Methods and Their Application. Cambridge, UK: Cambridge University Press.

11.

D’Hauteserre

(2000). “Lessons in Managed Destination Competitiveness: The Case of Foxwoods Casino Resort,” Tourism Management, 21 (1): 23-32.

12.

Diamantopoulos

Winklhofer

H. M.

(2001). “Index Construction with Formative Indicators: An Alternative to Scale Development.” Journal of Marketing Research, 38: 269-77.

13.

Dwyer

Forsyth

(1994). “Foreign Tourism Investment: Motivation and Impact.” Annals of Tourism Research, 21(3): 512-537

14.

Dwyer

Forsyth

Rao

(2000). “The Price Competitiveness of Travel and Tourism: A Comparison of 19 Destinations.” Tourism Management, 21 (1): 9-22.

15.

Dwyer

Kim

(2003). “Destination Competitiveness: Determinants and Indicators.” Current Issues in Tourism, 6 (5): 369-414.

16.

Dwyer

Mellor

Livaic

Edwards

Kim

(2004). “Attributes of Destination Competitiveness: A Factor Analysis.” Tourism Analysis, 9 (1): 91-101.

17.

Ennew

C. T.

Reed

G. V.

Binks

M. R.

(1993). “Importance-Performance Analysis and the Measurement of Service Quality.” European Journal of Marketing, 27 (2): 59-70.

18.

Enright

M. J.

Newton

(2004). “Tourism Destination Competitiveness: A Quantitative Approach.” Tourism Management, 25 (6): 777-88.

19.

Enright

M. J.

Newton

(2005). “Determinants of Tourism Destination Competitiveness in Asia Pacific: Comprehensiveness and Universality.” Journal of Travel Research, 43: 339-50.

20.

Fabrigar

L. R.

Wegener

D. T.

MacCallum

R. C.

Strahan

E. J.

(1999). “Evaluating the Use of Exploratory Factor Analysis in Psychological Research.” Psychological Methods, 4 (3): 272-99.

21.

Fletcher

Archer

B. H.

(eds.). (1991). The Development and Application of Multiplier Analysis. Vol. 3. Chichester: Wiley.

22.

Geisser

(1975). “The Redictive Sample Reuse Method with Applications.” Journal of the American Statistical Association, 70: 320-28.

23.

Gooroochurn

Sugiyarto

(2005). “Competitiveness Indicators in the Travel and Tourism Industry.” Tourism Economics, 11 (1): 25-43.

24.

Gooroochurn

Sugiyarto

(2004). “Measuring Competitiveness in the Tourism and Travel Industry.” In International Conference on Tourism Modelling and Competitiveness. Cyprus.

25.

Hair

J. F.

Black

Babin

Anderson

R. E.

Tatham

R. L.

(2010). Multivariate Data Analysis. 7th edition. Upper Saddle River, NJ: Prentice Hall.

26.

Hassan

(2000). “Determinants of Market Competitiveness in an Environmentally Sustainable Tourism Industry.” Journal of Travel Research, 38 (3): 239-45.

27.

Heath

(2003). Towards a Model to Enhance Africa’s Sustainable Tourism Competitiveness. Pretoria: Department of Tourism Management, University of Pretoria.

28.

Henseler

Ringle

C. M.

Sinkovics

R. R.

(2009). “The Use of Partial Least Squares Path Modelling in International Marketing.” New Challenges to International Marketing: Advances in International Marketing, 20: 277-320.

29.

Homburg

C. H.

Workman

J. P.

Krohmer

(1999). “Marketing’s Influence within the Firm.” Journal of Marketing, 63 (2): 1-17.

30.

Hulland

(1999). “Use of Partial Least Squares (PLS) in Strategic Management Research: A Review of Four Recent Studies.” Strategic Management Journal, 20: 195-204.

31.

Inskeep

(1991). Tourism Planning: An Integrated and Sustainable Development Approach. New York: Van Nostrand Reinhold.

32.

Inskeep

(1987). “Environment Planning for Tourism.” Annals of Tourism Research, 14 (1): 118-135.

33.

Ioannides

Debbage

Keith

. (1998). The Economic Geography of the Tourist Industry: A Supply-Side Analysis. New York: Routledge.

34.

Jarvis

C. B.

Mackenzie

S. B.

Podsakoff

P. M.

(2003). “A Critical Review of Construct Indicators and Measurement Model Misspecification in Marketing and Consumer Research.” Journal of Consumer Research, 30 (2): 199-218.

35.

Johnston

D. F.

(1998). “Towards a Comprehensive Quality of Life Index”. Social Indicators Research, 20: 473-96.

36.

Jöreskog

Wold

(1982). “The ML and PLS Techniques for Modelling with Latent Variables: Historical and Comparative Aspects.” In Systems under Indirect Observation, Vol. 1, edited by Jöreskog

Wold

Amsterdam: North-Holland, pp. 263-70.

37.

Khadaroo

Seetanah

(2007). “Transport Infrastructure and Tourism Development.” Annals of Tourism Research, 34 (4): 1021-32.

38.

Kozak

(2002). “Destination Benchmarking.” Annals of Tourism Research, 29 (2): 497-519.

39.

Kulendran

Wong

K. K. F.

(2005). “Modelling Seasonality in Tourism Forecasting.” Journal of Travel Research, 44( 2): 1-8.

40.

Lee

F.C.

(1994). “An exploratory study of international tourism demand from selected countries to Taiwan.” Unpublished master thesis, Virginian Polytechnic Institute and State University, Blacksburg, Virginia.

41.

Mangion

M.-L.

Durbarry

Sinclair

M. T.

(2005). “Tourism Competitiveness: Price and Quality.” Tourism Economics, 11 (1): 24-68.

42.

Mazanec

Wober

Zins

(2007). “Tourism Destination Competitiveness: From Definition to Explanation?” Journal of Travel Research, 46: 199-209.

43.

Middleton

V. T. C.

(1997). “Sustainable Tourism: A Marketing Perspective.” In Tourism Sustainability. Principles to Practice, edited by Stabler

M. J.

Wallingford: CABI, pp. 13-26.

44.

Nunnally

J. C.

(1978). Psychometric Theory. New York: McGraw-Hill.

45.

Nunnally

J. C.

Bernstein

I. H.

(1994). Psychometric Theory. 3rd edition. New York: McGraw-Hill.

46.

Olinsky

Chenb

Harlow

(2003). “The Comparative Efficacy of Imputation Methods for Missing Data in Structural Equation Modelling.” European Journal of Operational Research, 15 (1): 53-79.

47.

Porter

(1990). The Comparative Advantage of Nations. New York: Free Press.

48.

Quazi

(2007). “Economic Freedom and Foreign Direct Investment in East Asia.” Journal of the Asia Pacific Economy, 12 (3): 329-44.

49.

Ritchie

J. R.

Crouch

(2000). “The Competitive Destination: A Sustainability Perspective.” Tourism Management, 21 (1): 1-7.

50.

Ritchie

J. R.

Crouch

(2003). The Competitive Destination: A Sustainable Tourism Perspective. Wallingford: CABI.

51.

Ritchie

J. R.

Brent

G. I.

Crouch

Hudson

(2001). “Assessing the Role of Consumers in the Measurement of Destination Competitiveness and Sustainability.” Tourism Analysis, 5 (2-4): 69-76.

52.

Russell

D. W.

Kahn

Spoth

Altmaier

E. M.

(1998). “Analyzing Data from Experimental Studies: Illustration of a Latent Variable Structural Equation Modeling Approach.” Journal of Counseling Psychology, 45: 18-29.

53.

Shan

Wilson

(2001). “Causality between trade and tourism: Empirical evidence from China.” Journal of Applied Economics Letters, 8: 279-283.

54.

Sinclair

M. T.

(1998). “Tourism and Economic Development: A Survey.” Journal of Development Studies, 34 (5): 1-51.

55.

Song

Witt

S. F.

(2003). “Tourism Forecasting: The General-to-Specific Approach.” Journal of Travel Research, 42: 65-74.

56.

Stone

(1974). “Cross-Validatory Choice and Assessment of Statistical Predictions.” Journal of the Royal Statistical Society, 36 (2): 111-33.

57.

Tenenhaus

(1998). La Régression PLS: théorie et pratique. Paris: Technip.

58.

Tenenhaus

EspositoVinzi

Chatelin

Y. M.

Lauro

(2005). “PLS Path Modelling.” Computational Statistics and Data Analysis, 48: 159-205.

59.

Tsai

Song

Wong

K. F.

(2009). “Tourism and Hotel Competitiveness Research.” Journal of Travel and Tourism Marketing, 26 (5): 522-46.

60.

Williams

P. W.

Stewart

Larsen

(2012). “Toward an Agenda of High-Priority Tourism Research.” Journal of Travel Research, 51 (1): 3-11.

61.

Witt

S. F.

Witt

C.A.

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

62.

Wold

(1979). “Model Construction and Evaluation When Theoretical Knowledge Is Scarce: An Example of the Use of Partial Least Squares” (Technical report). Geneve: Université de Geneve.

63.

Wold

(1982). “Soft Modelling: The Basic Design and Some Extensions.” In Systems under Indirect Observation, Vol. 2, edited by Jöreskog

K. G.

Wold

Amsterdam: North-Holland, pp. 1-54.

64.

Wold

H. O.

(1985). “Partial Least Square.” Encyclopedia of Statistical Sciences, 6: 581-91.

65.

XLSTAT. (2011). XLSTAT-PLSPM module, XLSTAT software, Addinsoft, Paris.