Assessing Island Destination Service Quality—An Integrated Model through Co-creation of Destination E-image

Destination E-image and Its Relationship with Behavioural Intentions

The ‘destination image’ in the tourism marketing context is the complex psychological image of products and services formed in the minds of tourists and potential tourists. This can be a pre-visit destination image and post-visit destination image (Beerli & Martín, 2004b). The pre-visit destination image is the influencer of the potential visitors’ expectations. Such expectation formation towards a destination image is related to motivation, experience and other socio-demographic characteristics of individuals (Beerli & Martín, 2004a; Sharafuddin & Madhavan, 2020a). In addition, external factors such as marketing efforts by tourism authorities, travel agencies, tour operators, hotels, restaurants, influence marketers, referrals and word-of-mouth also influence the expectation formation. In recent years, since the rise of social media platforms and high-quality camera smartphones, the expectation formation of destination image in the minds of potential visitors is highly influenced by the digital marketing efforts of the tourism organizations and digital footprints of the earlier visitors. Such digital footprints vary from simple geo-tagged locations with visitor sentiments through emojis to 4k full-length videos with detailed experience. Walden-Schreiner et al. (2018) used user-generated spatial data to assess the visitor distribution and movement patterns in Hawaii Volcanoes National Park. It was evident from the research that the visitor’s movement pattern was near the natural environment and infrastructure. The researchers reported the similarity in the pattern of visitor movement and photos taken at geo-tagged locations. This also proved the influence of user-generated Flickr photos on the visitors’ movement and behaviour. Hence, it is obvious that the user-generated digital content influences the perception towards a destination and further influences similar activities of visitors during their visit. Though the photos taken by the visitors are influenced by the previously seen images of the destination; they all do not mean the same (Caton & Santos, 2008). Because Stepchenkova and Zhan (2013) compared the destination marketing organizations (DMO) projected image of the destination and tourists’ user-generated destination image and their study found that the user-generated content offers a more eye-for-detail of the tourists’ interests and it directly reflects the visitors’ perceptions. Such factual photographs form certain expectations in the minds of potential tourists, which further affects their perceptions during the moment of truth (Kim & Stepchenkova, 2015). In addition, Dinhopl and Gretzel (2016) addressed the growing dominance of individual representation in photographs taken by visitors at the destination. However, similar to other tourists who take photographs of the destination (Rakić & Chambers, 2012), self-representative tourists, with their selfie images, also construct the meaning of the destination through various self-indulged activities (Dinhopl & Gretzel, 2016). Månsson (2011) documented the cyclical relationship between destination-image production process and replicative consumption. From the above literature, it is clear that the digital footprints of visitors play a crucial role in continuously reshaping the destination image. Since it is replicative in nature, the activities and images take snowball effects, and the positive destination image will further lead to high expectations whereas a negative destination image will lead to low expectations which will further affect the perception during the visit. Hence, the overall service quality of the destination is one of the crucial elements reshaping the destination image. Though there are several computer science-related studies for assessing the user generated destination image (Bastidas-Manzano et al., 2021; Cao et al., 2020; Marine-Roig & Anton Clavé, 2015), research on the psychological factors that drive the psychical destination image formation through digital user-generated contents is relatively new. A recent study by Wang et al. (2021) found that tourist perception on destination image changes during on-site visit and post-visit through the specific process of each tourist visit. Therefore, in this research, we adopt the keyword ‘destination e-image’ to mention the digital destination image formed through various sources, including user-generated content in various social media platforms such as Facebook, Twitter, Instagram, Flickr and other digital platforms. Hence, the destination e-image formed by the visitors’ user generated contents acts as a strong digital-word-of-mouth influencer in potential visitors’ decision-making choices. Therefore, the following hypothesis was formed.

Destination E-image, Destination Service Quality and Destination Environmental Competitiveness

The destination e-image influences the destination service-quality expectations of potential visitors. If the user-generated-contents of the previous visitors project a high quality of environment and services, the potential visitors surfing and gaining knowledge of earlier visitors’ experiences from the internet will also develop similar expectations for their visit. But the quality perception of the destination’s natural attractions and services also significantly differ based on the tourists’ social-cultural factors, such as country of origin (Dedeoğlu et al., 2020). In tourism marketing literature, the concept of service quality is inspired by the early marketing literature and the conceptual service quality models of Parasuraman et al. (1985), Zeithaml et al. (1996). Babakus and Boller (1992) suggested that the service quality dimensions depends on the type of service industry and service under study. Hence, the variables selected, and statistical tools used within the tourism industry vary due to the dynamic nature of the industry. There are several research conducted in the past to assess sectors within the travel and tourism industry, such as SERVQUAL for airlines (Chou et al., 2011; Tsaura et al., 2002), hotels (Akbaba, 2006; Wilkins et al., 2009), restaurants (Ryu et al., 2012; Stevens et al., 1995), ECOSERV for ecotourism (Khan, 2003) and tour operations (Hudson et al., 2004). However, potential tourists will view the destination from their consumption point of view. Hence, various factors such as individual stimuli, features, facilities and overall attributes of the destination influence the pre-purchase destination selection process of potential tourists (Sharafuddin & Madhavan, 2020a). Thus, the consumption of services at a destination is not limited to one particular product or service but the essential tourism-related services of the whole tourist destination itself. However, limited research is available from the potential tourists’ consumption point of view. Wang et al. (2007) adopted a multi-stage approach for assessing the quality of group tour packages with ‘transportation’, ‘hotel accommodation’, ‘shopping arrangement’, ‘optional tour’, ‘tour leader’ and ‘local guide’ as variables. Tosun et al. (2015) adopted the ‘destination benchmarking’ concept of Kozak and Rimmington (1998) and the ‘destination product’ concept of Murphy et al. (2000) to develop the ‘destination quality’ model. They argued that all organizations in a destination offering tourism services are ‘interlocked’ and together form the destination service quality and thus measured the destination quality with ‘destination service quality’ (‘accommodation’, ‘local transport’, ‘cleanliness’, ‘hospitality’, ‘activities’, ‘language’ & ‘airport’) and ‘destination natural quality’ (‘location’, ‘culture’ & ‘pure beauty’). Several researchers viewed the natural quality of nature-based tourist destinations as the major element of destination competitiveness (Hu & Wall, 2005; Lundie et al., 2007; Mihalič, 2000; Murphy et al., 2000). Because a well-maintained sustainable natural environment will lead to tourist satisfaction and further lead to long-term positive appeal for selected target market segments (Hassan, 2000). Geographical research areas like the one in this research, that is, small islands, are unique in the environment, and the majority of the service providers are MSMEs (micro, small and medium-sized enterprises) in nature. Haven-Tang and Jones (2005) summarized the coherence between the ‘public-sector interventions’, ‘tourism SME (small and medium enterprise) approaches’ and ‘destination coherence.’ Other than the periodic approaches such as framing policies for operations, quality, standards, budgeting, funding, human resource development and skill training, visitor-based approaches such as marketing and market intelligence are crucial to take advantage of the full potential of the near real-time situation and maximize the destination competitiveness; especially, the public-private interventions in managing the destination e-image (Arlt, 2005; Getz et al., 2005). However, the visitors’ perception of the quality of tourism SME services is significantly different from that of large enterprises such as five-star chain hotels (Michael Hall & Rusher, 2005). Service providers’ knowledge and skills’; ‘business behaviours’ such as personalized sales and service; and ‘service attitude’ were the pre-dominant factors determining tourism SME service quality (Jones & Haven-Tang, 2005), which will collectively lead to destination service quality. Surprisingly, there is very limited research which discuss the relationship between the quality of services offered on an island and the destination environmental competitiveness. Mechinda et al. (2010) investigated the competitiveness factors of Koh Chang Island in Thailand and found that both quality of service and natural resources significantly influence tourist loyalty. However, the causal relationship between the destination environmental competitiveness and island destination service quality has not yet been studied.

Therefore, the construct for assessing the destination e-image, destination service quality and destination environmental competitiveness was adopted from the above literature, and the following hypotheses were formed:

Maritime Passenger Transport Service Quality (MPTSQ)

Maritime passenger transport systems such as ferry services play a vital role in island destination development. They are also an integral part of passenger transport systems in Southeast Asian countries and other archipelago regions. One of the earliest studies in assessing the service quality of maritime passenger transport services with special reference to ferries can be traced back to Pantouvakis (2007). The author assessed the service quality of ferry services using (a) Quality of Service (‘Confidence’, ‘Empathy’, ‘Viability’, ‘Politeness’, ‘Safety’, ‘Local’, ‘Speed’) and (b) Convenience (‘Departure’, ‘Arrival’), (c) Price (‘Cheap’, ‘Value’) and (d) ‘Availability’ and found them to be the determinants of satisfaction. Mathisen and Solvoll (2010) assessed the maritime passenger-transport service quality using 16 service elements such as (a) ‘frequency’, (b) ‘opening hours’, (c) ‘Schedule’, (d) ‘Fare’, (e) ‘Discount’, (f) ‘Regularity’, (g) ‘Punctuality’, (h) ‘Capacity Summer’, (i) ‘Capacity Winter’, (j) ‘Ferry Size’, (k) ‘Comfort’, (l) ‘Speed’, (m) ‘Information’, (n) ‘Service’, (o) ‘Cleaning’ and (p) ‘Catering’. They segmented the respondents into enterprises, household commuters, Business travellers and leisure travellers and found that comparatively, leisure passengers gave less importance to all the service elements. However, they were more satisfied with the capacity. Tanko et al. (2019) assessed the satisfaction of passengers using an urban water transit system. The authors used three attributes and related variables as; (a) service (‘frequency’, ‘network’, ‘access’, ‘punctuality’), (b) comfort (‘cleanliness’, ‘calmness’, ‘open marine environment’, ‘view from the boat’) and (c) productivity (‘ability to work’, ‘space to work’, ‘smoothness of the ride to work’) and found that the comfort attributes were more appropriate in assessing the passenger satisfaction. However, those studies were more precise towards the core operations of the vessel but did not include the services such as ease of booking, waiting area and jetty services. Hart et al. (2020) assessed the ferry services using ‘jetties facilities’, ‘Fare’, ‘Safety & security’ and ‘comfort’ and found that jetty infrastructure, ease of booking online, ambience and on board convenience were the most important determining factors of service quality. Chan (2017) adopted the modified SERVQUAL model of Cavana et al. (2007) with ‘Assurance,’ ‘Empathy,’ ‘Reliability,’ ‘Responsiveness,’ ‘tangibles,’ ‘Comfort,’ ‘Connection,’ and ‘Convenience’ as dimensions to assess the passenger ferry service quality and found it to be suitable for identifying the most and least important factors of maritime passenger transport service quality (MPTSQ). Abdullah and Wahab (2010) used Fuzzy multi-criteria Decision-making approach with (a) On-board Comfort (Cleanliness & Noise level of ferry, On-board facilities), (b) Ferry Employees (Helpful attitude, attentiveness, service efficiency) and (c) Handling of abnormal conditions (Security, Safety, and on-time performance) and found service efficiency as the highest-ranked attribute. Marissa et al. (2019) built a scale with 12 vessel-related variables and 32 port-related variables to assess five satisfaction-related variables. There is a strong relationship between maritime passenger transport service quality, tourist satisfaction and behavioural intentions. (Dewaayusastagayatri Dewi & Respati, 2020). But the contribution of MPTSQ to the overall IDSQ in not studied separately. Also, the mediating role of IDSQ on the relationship between MPTSQ and TS and the mediating role of TS in the relationship between MPTSQ and BI remain unexplored. In addition, the role of destination e-image in visitor’s perception formation towards the maritime passenger transport system also remains unreported in academic research. Therefore, the construct for assessing the MPTSQ was adopted from the above literature, and the following hypotheses were formed.

Destination Service Quality, Tourist Satisfaction and Behavioural Intentions

The influential relationships between service quality, tourists’ satisfaction towards the destination and their behavioural intention is well documented (Prayogo & Kusumawardhani, 2016; Stylos et al., 2017; Whang et al., 2016). However, the influential relationship between these three is not limited to individual entities such as hotels and restaurants. It also includes the activities undertaken by the visitors at the destination (Sharafuddin, 2015a). Sangpikul (2018) investigated the destination attributes as dimensions of travel experience (beach attractions, people, valued destination, services & facilities, safety & cleanliness) and their effects on tourist satisfaction and behavioural intention. The research found that the people (local hospitality of service providers) other than the natural environment also influences the tourist satisfaction and behavioural intentions in island destinations. Several studies (Fernaldi & Sukresna, 2018; Majeed et al., 2020; Prayag et al., 2013; Vega-Vázquez et al., 2017) also proved the mediating effects of tourist satisfaction between the tourist experience and behavioural intention. Hence, the following hypotheses were formed based on the above literature.

Conceptual Framework

The following conceptual framework was derived from the qualitative study (Sharafuddin et al., 2021) and literature review, and the following hypotheses were formed (Figure 1).

OPEN IN VIEWER

Figure 1.

Source: Authors’ own conceptualization.

Scale Development

This research adopted an empirical approach by using a survey instrument to investigate the causal impacts among destination e-image (DEI), maritime passenger transport service quality (MPTSQ), island destination service quality (IDSQ), destination environmental competitiveness (DEC), tourist satisfaction (TS) and behavioural intentions (BI). The survey instrument was designed based on the qualitative study (Sharafuddin et al., 2021) and literature review, which was subsequently discussed with the three subject-matter experts. The survey instrument is comprised of two sections; the first section consisted of questions related to the demographic profile of the visitors, and the second section intended to measure the perception of visitors related to their experience and satisfaction with the destination. The survey instrument consisted of 36 items on a 5-point Likert scale ranging from 5-strongly agree to 1-strongly disagree. The scale items are related to destination e-image (7 items), maritime passenger transport service quality (11 items), island destination service quality (6 items), destination environmental competitiveness (4 items), tourist satisfaction (5 items), and behavioural intentions (3 items). The survey instrument was presented again to the same three subject-matter experts to evaluate the item constructs in terms of five aspects, that is, research domain, research theme, its importance to the concept, suitability and compatibility to the study’s purpose. The experts suggested changing the wordy sentences to ease the understanding of the respondents, and subsequent paraphrasing was done to the few questions; thus, it helped to achieve and ensure the content validity of the survey instrument. Thus, the final survey instrument consisted of 36 items.

Data Collection

This study included only repeat visitors to test the cyclical influence of their behavioural intention and destination e-image formation. The survey instrument (Refer to Annexure 1) was distributed to the repeat visitors visiting Koh Larn Island in December 2020 using the convenience sampling method. Both ‘paper-pen’ approach and an online survey form were used to collect data in Bali Hai Pier (Pattaya City) from the visitors returning from Koh Larn Island. The survey instrument was distributed to 600 respondents; out of 600 respondents, 542 respondents filled the survey forms. In 542 responses, 158 responses were partially incomplete/day-time visitors and hence omitted from the study. Thus, the remaining 384 complete responses were included in the study. With a 5% margin error and 95% confidence level and 50% response distribution, the minimum sample size required from a population size of more than 20,000 is 377 (Raosoft, n.d.). Also, as a rule-of-thumb (Comrey & Lee, 1992) suggested that 300 cases are good for conducting factor analysis, whereas on the other hand, the minimum sample size based on the number of variables was suggested by (Gorsuch, 1983), that is, at least five cases per variable. The author (Kahn, 2006) suggested that the sample of 300 cases is sufficient and safe to have communalities in an appropriate range for a good solution. Russell (2002) stated that the sample size should be greater than 100 to conduct a confirmatory factor analysis (CFA). MacCallum et al. (1999) has recommended that a higher sample size would lead to less sampling error and reduce items’ misclassification into different factors. Hence, the sample size of 384 is sufficient to test the hypothesis. The confirmatory factor analysis was conducted to test the validity and reliability of the survey instrument. Consequently, the causal impact was measured using the structural equation model. The EFA, CFA and SEM model was performed using the statistical programming language R (R Core Team, 2020) and the Lavaan package (Rosseel, 2012).

Scale Reliability and Validity

The scale reliability was assessed for all the six factors (destination e-image, maritime passenger transport service quality, island destination service quality, destination environmental competitiveness, tourist satisfaction, and behavioural intentions) with 36 items. The internal consistency and reliability can be assessed through the score of Cronbach’s alpha and corrected item-total correlations. Table 1 presents the values of Cronbach’s alpha and corrected item-total correlation (CITC). The Cronbach’s alpha values are greater than the threshold value of 0.70 (Nunnally, 1978; Nunnally & Bernstein, 1994), indicating good reliability; and the CITC values of all the items were above 0.50 (Iacobucci & Churchill, 2010) except one item in DEI; thus it (DEI1) was subsequently removed from the model.

OPEN IN VIEWER

Table 1.

Reliability Test

S. No.	Constructs	Number of Items	Cronbach’s Alpha	Corrected Item-total Correlation Range
1	Destination e-Image	6	0.852	0.545–0.676
2	Maritime passenger transport service quality	11	0.907	0.530–0.714
3	Island destination service quality	6	0.863	0.610–0.721
4	Destination environmental competitiveness	4	0.849	0.630–0.729
5	Tourist satisfaction	5	0.848	0.564–0.746
6	Behavioral intentions	3	0.861	0.710–0.781

Source: The authors.

The number of factors for EFA was estimated as six based on the conceptual framework and scree plot (Costello & Osborne, 2005). Further, 384 responses to 35 items in the survey instrument were subjected to exploratory factor analysis (EFA) with principal axis factoring and promax rotation. Researchers in the field of social sciences have suggested that principal axis factoring (PAF) extraction with promax/direct oblimin rotation is more appropriate (Brenner, 2019), and PAF extraction is more accurate than principal component analysis (Kahn, 2006). The communalities extracted for all the 35 items ranged from 0.420 to 0.696, which is greater than 0.40 and acceptable in social sciences (Costello & Osborne, 2005). The factor model accounted for 56.39% of the variance explained, which is greater than 50% (Streiner, 1994). The first factor accounted for 30% of the total variance. The KMO measure of sampling adequacy score was shown at 0.897, which is also greater than 0.50 (Kaiser, 1974). However, few cross-loadings of items were found as we used Promax rotation and allowed them to correlate as there are possibilities for the items to be correlated in a related construct. Thus, the cross-loadings were verified and the items MPTSQ1, MPTSQ2, MPTSQ3, TS1 and TS5 were loaded high in another factor, so these items have been subsequently removed. The exploratory factor analysis was conducted again using principal axis factoring and Promax rotation with six factors. The factor model accounted for 56.887% of the total variance explained, which is also greater than 50% (Streiner, 1994). The first factor accounted only for 29.918% of the total variance, indicating that most of the variance is not accounted by one general factor; hence, the common variance problem was addressed and alleviated in this study (Podsakoff et al., 2003). The KMO measure of sampling adequacy score was shown at 0.890, which is also greater than 0.50 (Kaiser, 1974). Hence, to confirm the factor structure and to verify the discriminant validity, the final 30 items were subjected to confirmatory factor analysis.

The initial measurement model was assessed for convergent validity and discriminant validity using confirmatory factor analysis with all 30 items of the scale construct. The measurement model was converged with the Chi-square value of χ² = 1178.846 with degrees of freedom (df) 390, p ≤ 0.01, where root mean square approximation (RMSEA) = 0.073. The factor loadings of all items were > 0.5, with z-values ranging from 11.038 to 20.139 and a significant p < 0.05. Though, the model yielded required factor loadings with significant z-values, the model achieved only the marginal fit (CMIN/DF = 3.02, RMSEA 0.073, standardized root mean square residual [SRMR] = 0.058, GFI = 0.83, CFI = 0.872, TLI, 0.858, NNFI = 0.858, NFI = 0.822, IFI = 0.873). The model fit indices indicate that the model could be improved further by inspecting the modification indices and standardized residuals to identify the potential items for deletion (Pattnaik, 2019). The modification indices (MIs) were carefully reviewed by giving due consideration to theory, as MIs are completely data-driven (CenterStat, 2019). The MIs and standardized residuals pointed out the items/pairs with high correlated error terms. These items were prudently checked from a theoretical perspective, including language, wordiness, and its suitability to the present pandemic situation, so that there will not be much changes to the hypothesized model. The error items MPTSQ8, DSQ2, DSQ5, DEI3, DEI6, DEC4 were dropped and revised in the model based on the recommendations of Hair et al. (2010). The final model was converged with the Chi-square value of χ² = 559.187 with degrees of freedom (df) 237, p ≤ 0.05, where root mean square approximation (RMSEA) = 0.059. The R² values are greater than 0.25 with factor loadings of all items were > 0.5 (Hair et al., 2010), and z-values ranged from 10.905 to 19.991 with a significant p-value of <0.05 (Table 2). The model yielded required factor loadings with significant z-values and achieved a good fit (CMIN/DF = 2.35<3, RMSEA = 0.059, RMR = 0.030, SRMR = 0.048, GFI = 0.90, CFI = 0.92, TLI = 0.91, NNFI = 0.91, NFI = 0.90, IFI = 0.93), which is presented in Table 4. Further, the data were tested to check the risk of common method bias. Common method bias such as ‘transient mood state’ may occur while collecting data from tourists because there is a possibility of the respondents’ previous moods to affect their responses. Also, there are possibilities of ‘consistency motif’ and ‘leniency biases’ in research where respondents are asked to fill in their experience and behavioural intention. Therefore, the data were tested for the effects of common method bias. Harman’s single factor test is the one-factor exploratory analysis of all the variables used in the study to identify the presence of common method bias in the research data. The results of the proportion variance from the factor analysis should be less than 0.50 (50%) (Harman, 1967). Therefore, the single factor exploratory analysis was carried out with all the 24 variables included in this study to test the hypothesis. The proportion variance of the single-factor analysis result was 0.30 (30%). Hence, the common method bias does not affect the data.

OPEN IN VIEWER

Table 2.

Results of Measurement Model

	Results of Confirmatory Factor Analysis			Reliability Test Results
Variables	Standard Estimates	z-Value	R 2	Composite Reliability
Destination e-image				0.79
DEI2: My social network and user-generated social media contents.	0.564	10.905	0.318
DEI4: Hotels commercial contents/emails.	0.687	13.890	0.473
DEI5: Online tourist review contents.	0.788	16.481	0.621
DEI7: My own experience and digital contents.	0.716	14.611	0.513
Maritime passenger transport service quality				0.88
MPTSQ4: The ferry frequency is adequate.	0.650	13.629	0.422
MPTSQ5: The ferry arrival/departure service is punctual, and I can rely on the time.	0.755	16.678	0.570
MPTSQ6: I feel at ease of boarding and alighting.	0.752	16.586	0.565
MPTSQ7: The ferry seats are comfortable, and the boat is clean.	0.767	17.049	0.588
MPTSQ9: I feel comfortable with the ride safety, convenience and speed.	0.748	16.472	0.560
MPTSQ10: The manners of ferry staff on-board are polite.	0.696	14.920	0.485
MPTSQ11: The trip is value for money.	0.663	13.991	0.440
Island destination service quality				0.81
IDSQ1: The accommodation services are good.	0.731	15.451	0.535
IDSQ3: The island looks clean and tidy.	0.747	15.893	0.558
IDSQ4: The hospitality of tourism SMEs (Small and Medium Enterprise) (Motorbike rentals, food shops, restaurants) is good.	0.677	13.954	0.458
IDSQ6: The level of services in the island and marine activities I involved during the trip were at par.	0.714	14.967	0.510
Destination environmental competitiveness				0.81
DEC1: The beaches are clean and litter free.	0.741	15.754	0.549
DEC2: The seawater is clear.	0.818	17.980	0.669
DEC3: The marine ecosystem is lively (can see marine life like corals, small fishes).	0.734	15.545	0.538
Tourist satisfaction				0.78
TS2: The social distancing practices of the service providers are satisfactory.	0.654	13.098	0.428
TS3: The social distancing practices by other visitors are satisfactory.	0.759	15.722	0.576
TS4: Overall, my travel experience with the destination is good.	0.796	16.687	0.634
Behavioral intentions				0.87
BI1: Intention to recommend others using online platforms.	0.815	18.534	0.664
BI2: Intention to share the experience online.	0.858	19.991	0.736
BI3: Intention to revisit.	0.803	18.141	0.644

Source: The authors.

Composite Reliability, Convergent Validity and Discriminant Validity

The composite reliability of the latent constructs is depicted in Table 2. The composite validity of the latent constructs was obtained through confirmatory factor analysis and the observed composite reliability (CR) values ranging from 0.78 to 0.88 (which is greater than 0.7), indicated good reliability, and the factor loadings of few items between >0.5 and <0.7 does not affect the internal consistency of the scale constructs (Hair et al., 2010). The CR values greater than 0.70 are considered as an indicator of convergent validity, and it is evident that the measurement scale met the criteria of convergent validity (Table 2), that is, composite reliability >0.7 (Hair et al., 2010). The idea of discriminant validity was prompted by Campbell and Fiske (1959) and a few other experts in the field. The discriminant validity measures the ‘similarity between latent variables’ (Henseler, n.d.). The discriminant validity was assessed using the novel and stringent measure Heterotrait-Monotrait ratio of correlations (HTMT) (Henseler et al., 2015) in R studio using Lavaan as the traditional Fornell and Larcker criterion (Fornell & Larcker, 1981) is purely criterion-based and not based on the statistical test (Henseler et al., 2015). Table 3 presents the HTMT results of the six latent variables and demonstrates the evidence of discriminant validity where the HTMT values between the latent variables are <0.85, according to the HTMT_0.85 criterion (Henseler et al., 2015).

OPEN IN VIEWER

Table 3.

HTMT Results Discriminant Validity

Latent Variables	DEI	MPTSQ	DSQ	DEC	TS	BI
DEI	1	0.382	0.508	0.485	0.324	0.517
MPTSQ	0.382	1	0.485	0.298	0.302	0.320
DSQ	0.508	0.485	1	0.676	0.553	0.611
DEC	0.485	0.298	0.676	1	0.567	0.710
TS	0.324	0.302	0.553	0.567	1	0.639
BI	0.517	0.320	0.611	0.710	0.639	1

Source: The authors.

Table 4 presents the obtained value and threshold level of model fit indices. In general, the researchers consider other indices as the chi-square value is sensitive to sample size (Bentler & Bonett, 1980); hence other fit indices should be considered for assessing the model fit (Hooper et al., 2008). According to Hair et al. (2010), as a rule of thumb, if at least one absolute and incremental fit index meets the threshold level apart from the Chi-square test statistic, the model is said to be a good fit. Thus, from Table 4 it is evident that the measurement model satisfied the threshold levels of major indices, which indicates a good fit.

OPEN IN VIEWER

Table 4.

Model Fit Indices

Fit Index	Threshold Level	Obtained Value	Reference
Chi-Square (χ2)	p > 0.05	559.187 with df 237 and p < 0.05	Insignificant p-value rarely occurs in CFA as Chi-square value is sensitive towards sample size (Bentler & Bonett, 1980)
Absolute fit indices
CMIN/DF	<3	2.35	Good fit (Hu & Bentler, 1999)
Goodness of fit index (GFI)	≥0.90	0.90	Good fit (Ahire et al., 1996; Hooper et al., 2008; Jöreskog, 1984)
Root mean square error of approximation (RMSEA)	<0.08	0.059	Good fit (Hu & Bentler, 1999; MacCallum et al., 1996)
Root mean square residual (RMR)	<0.09	0.030	Good fit (Hu & Bentler, 1999)
Standardized root mean square residual (SRMR)	<0.08	0.048	Excellent Fit (Hu & Bentler, 1999
Incremental fit indices
Comparative fit index (CFI)	>0.90	0.92	Good fit (Hair et al., 2010; Hu & Bentler, 2009; Pan et al., 2017; Tanaka & Huba, 1985)
Tucker-Lewis index (TLI)/Bentler-Bonett non-normed fit index (NNFI)	>0.90	0.91	Good fit (Bentler & Bonett, 1980; Franke & Byrne, 1995; Pan et al., 2017; Tanaka & Huba, 1985)
Bentler-Bonett normed fit index (NFI)	>0.90	0.90	Acceptable fit (Forza & Filippini, 1998; Kim et al., 2016)
Bollen’s incremental fit index (IFI)	>0.90	0.93	Good fit (Bollen, 2016; Henry & Stone, 1994)
Relative non-centrality index (RNI)	Close to 0.95	0.92	Good fit (Hu & Bentler, 1998)
Bollen’s relative fit index (RFI)	Values close to 1	0.86	Good fit (Hu & Bentler, 1999)

Source: The authors.

Causal Diagram

Overall Fitness of the Model

The final model was converged with the Chi-square value of χ² = 566.345 with degrees of freedom (df) 239, p ≤ 0.05. The z-values range from 9.493 to 18.196 with a significant p-value of <0.05. The model yielded required factor loadings with significant z-values and achieved a good fit (CMIN/DF = 2.36 < 3 (Hu & Bentler, 1999), RMSEA = 0.06 < 0.08 (MacCallum et al., 1996), RMR = 0.03 < 0.08 (Hu & Bentler, 1999), Standardized root mean square residual (SRMR) = 0.05 < 0.08 (Hu & Bentler, 1999), GFI = 0.90 (Ahire et al., 1996), CFI = 0.92 > 0.90 (Bentler, 1990; Hair et al., 2010; 2013; Pan et al., 2017), TLI = 0.91 > 0.90 (Bentler, 1990; Bentler & Bonnett, 1980; Byrne, 1994; Pan et al., 2017; Tanaka & Huba, 1985), IFI = 0.92 > 0.90 (Henry & Stone, 1994). The model achieved the suggested threshold levels.