Dynamic capabilities in tourism businesses: antecedents and outcomes

‘Dynamic capabilities’ are defined in the literature in various ways. Some scholars perceive them as underpinning processes and routines that facilitate the development of firms (Eisenhardt and Martin 2000; Winter 2003; Zott 2003). Others view operational issues as existing at multiple levels of the organization, such as operating portfolios of projects which require several operational capabilities to achieve short-term and long-term objectives (Dasari et al. 2015). Some construe them as activities (Ali et al. 2012; Teece 2014). A further group of scholars views dynamic capabilities as resources, assets, or the organization itself. In this study, we agree with Teece’s (2018) view regarding the nature of dynamic capabilities. Teece (2018) contends they are not resources, which thus need to be distinguished from dynamic capabilities even though the dynamic capabilities view (DCV) is an extension of the resource-based view (RBV). This view avoids confusion with the organizational assets or resources that form the core of dynamic capabilities. Our study adopts the definition of dynamic capabilities originally suggested by Teece et al. (1997, p. 516) as “the firm’s ability to integrate, build, and reconfigure internal and external competencies to address rapidly changing environments.”

In tourism research, DCV is applied as the theoretical framework in a number of research projects. Most research focuses on the hotel sector (Fraj et al. 2015; Leonidou et al. 2015; Marco-Lajara et al. 2017). Others focus on tourism innovation (Sainaghi et al. 2017; Verreynne et al. 2019) and competitive advantage (Evans 2016; Nieves and Haller 2014), especially eco (Leonidou et al. 2015) and green competitive advantage and management (Mittal and Dhar 2016) in tourism SMEs (Brida et al. 2016). Most research utilizes DCV as a theoretical lens to view and explain competitive advantage, the sustainability of competitive advantage, performance, or firms' innovation. As such, little research analyzes how dynamic capabilities and their micro-foundations are manifested in tourism firms.

Dynamic capabilities can be viewed as a combination of all capabilities relating to sensing, seizing, and reconfiguring/transforming and not as a single independent capability of each (Nyamrunda and Freeman 2021). According to Byrne (2010), the higher-order model represents a seemingly distinct hypothesis, and associated constructs can be gauged by one or more common underlying higher-order construct(s). Furthermore, the second-order factor model is more parsimonious and provides error-free estimates of both general and specific factors (Chen et al. 2005). Therefore, we argue that a second-order (higher-order) factor model is more precise than a first-order model in measuring dynamic capabilities. In addition, the higher-order three-factor model of dynamic capabilities provides insights into how the three first-order factors contribute to overall dynamic capabilities.

Discussions on the outcomes of dynamic capabilities focus on improving organizational performance (Augier and Teece 2009; Wilden et al. 2013). When evaluating this relationship to performance, numerous criteria and indexes can be used, of which the two most common are the financial performance and strategic performance of organizations in gaining a desirable position in markets to achieve the set goals. Competitive advantage is regarded as a “superior market position” (Weerawardena 2003, p. 21) that organizations possess to provide the market with superior products or services while attaining market dominance (Hunt and Morgan 1995). Previously, a firm’s primary objective was to achieve superior financial performance. Financial indicators were considered the sole and vital measures of an organization’s performance (Barney 1991). Financial performance is often specified by indicators such as profits and return on investment (Hunt and Morgan 1995). However, the firm's success cannot rely on using exceptional financial performance as a sole indicator of competitive advantage (Barney 1991; Day and Wensley 1988). From an RBV, the organization attains superior financial performance when it has superior skills or resources (Arbelo et al. 2020). This results in positional advantages and superior performance outcomes regarding relative profits and market share (Seyoum 2020).

Consequently, by nature, competition involves constant attempts by firms to obtain a comparative advantage in the resources they possess, which will yield a marketplace position of competitive advantage (Hunt and Morgan 1996). This will result in superior financial performance (Hunt and Morgan 1995). Porter (1991, p. 96) argues that a firm succeeds when it gains a “superior and sustainable performance… relative to the world’s best rivals”. Therefore, in evaluating the competitive position of an organization, it is necessary to look at both financial performance and marketplace position. This study thus defines ‘competitive advantage’ as the ‘superior market position’ an organization possesses to provide the market with superior products and services. These superior products and services then result in a positional advantage and superior performance outcomes with respect to profits and market dominance (Hunt and Morgan 1995; Jantunen et al. 2018; Porter 1991).

Research on the tourism industry discusses competitive advantage from the perspective of the RBV (Liu 2017; Molina-Azorin et al. 2010). The different resources for competitive advantage from the RBV include knowledge (Fraj et al. 2015; Thomas and Wood 2014), leadership capabilities (Sainaghi et al. 2017), innovation capabilities (Kale et al. 2019), and/or location (Denicolai et al. 2010; Molina-Azorin et al. 2010). Recent research by Verreynne et al. (2019) focuses more heavily on green competitive advantage and sustainable competitive advantage. There is a strong and compelling argument by Evans (2016) on how to facilitate innovation for sustainable competitive advantage.

Dynamic capabilities are important for tourism firms because the industry is now facing changes in all aspects of the service process, with rapid technological digitalization as the main factor. Some examples of technological advancements that are revolutionizing the industry are mobile technology, big data, the Internet of Things, blockchain, 5G technology, and augmented reality, to name just a few. The transformation that tourism companies need to focus on is not merely on changing the travellers’ experience but also on reforming the whole system, leading to significant innovation or even the creation of new business models. Such changes in technology and market trends require companies operating in the tourism industry to develop the dynamic capabilities to rapidly identify, transform, reform business processes, or create new operation models. All these activities will require ordinary capabilities needed for daily operations and using existing resources (Qaiyum and Wang 2018; Schriber and Löwstedt 2020) and superior capabilities such as dynamic capabilities to identify resources to execute the innovative initiatives at a higher level. As such, Nieves et al. (2016) argue that innovation in tourism services is based on capabilities for developing knowledge and learning and that the dynamic capabilities approach is a valuable framework for investigating innovation in the industry (Camisón and Monfort-Mir 2012). Innovation in tourism is also suggested on the basis of service design and service processes, rather than merely product innovation (Gomezelj 2016).

By reconfiguring current resources, organizations that possess dynamic capabilities will have a more positive impact on the firm's competitive advantage (Eisenhardt and Martin 2000; Teece and Pisano 1994). In a study by Schilke (2014), the influence of dynamic capabilities on competitive advantage was found to be contingent on the level of external dynamism. Dynamic capabilities are considered key to acquiring a competitive advantage in strategic management (Agwunobi and Osborne 2016; Li and Liu 2014). They are also the critical antecedents driving the innovation-based competitive advantage process (Chatzoglou and Chatzoudes 2018; Muhic and Bengtsson 2019; Salunke et al. 2019), determinants for achieving a competitive advantage (Nieves and Haller 2014), and a handy tool for organizations to gain a competitive advantage, even in a highly volatile environment (Monteiro et al. 2019; Schwarz et al. 2020). Therefore, the following hypothesis is proposed:

Figure 1 summarises the relationships between dynamic capabilities and their antecedents and outcomes.

Vietnam is used as the research context for this paper. It is one of Southeast Asia’s most picturesque countries, attracting an increasing number of travelers to its lush mountains, dynamic cities, and sandy beaches before COVID-19. Tourism in Vietnam has a vast potential for development, and it remains an attractive destination despite the pandemic challenges.¹ The World Economic Forum (2019) places Vietnam 63rd among 140 economies for tourism competitiveness, one of the top ten most improved economies, and the fastest improving nation in the Association of South East Asian Nations (ASEAN) (Calderwood and Soshkin 2019). According to the World Economic Forum (2017), Vietnam is predicted to be one of the top ten destinations for travel from 2016 to 2026 (Crotti and Misrahi 2017). These figures demonstrate the rapid and dynamic growth of the tourism sector in Vietnam and the opportunities it provides.

The research population was determined as Vietnamese domestic tourism companies. According to official figures from the Vietnam National Administration of Tourism, in 2017, there were 824 hospitality businesses from 3 stars and above and 1430 tourism companies nationwide (VNAT 2018). The sample size was determined as per the rule of thumb suggested in multivariate analysis literature (Hair et al. 2014). Specifically, there are 42 items developed for our measurement model. According to the criterion suggested by Hair et al. (2014) for choosing the expected sample size, a subjects-to-items ratio should be from 5 to 20 to ensure possible identification of the proposed statistical model. This means a sample N ranging from 210 to 840 was acceptable for running CFA and SEM in this study. Given the time constraints and financial resources for data collection, 1,000 questionnaires were distributed with a usable target response rate of at least 21% (to yield at least N = 210). After the sampling criteria, target population and sample were identified, this research utilized different sampling methods to optimize the response rate.

The sampling techniques used to collect data consist of convenient sampling, stratified sampling, and snowball sampling to represent different types of businesses in the tourism sector and represent the three key regions of Vietnam (the North, the Centre, and the South). The data collection was undertaken in three stages. The first stage utilized convenient sampling conducted at the national and international tourism fairs in Vietnam's two biggest cities: Ho Chi Minh City (in March 2017) and Hanoi City (in April 2017). This technique generated 112 responses from both events.

The second stage involved stratified sampling whereby the number and contact details of travel and hospitality businesses were taken from the Vietnam National Agency for Tourism (VNAT) statistics section. The percentage of each province’s businesses over the total national number (1,578 travel companies and 1,577 hospitality addresses across 63 provinces) was calculated, and the subsequent numbers of tourism firms required in those provinces were determined. Resource constraints meant that half this number would be required. The calculation showed that 500 questionnaires should be distributed to 53 representative provinces (as there are ten provinces with less than 0%). Therefore, 500 pre-paid postal questionnaires were sent and administered from March to May 2017. This technique did not generate a reasonable response rate even with follow-up calls where possible. Only 10 responses were returned out of 500 distributed questionnaires.

During the first and second stages of sampling, it became apparent that the response process was prolonged and limited. Therefore, we decided to initiate the third stage combined with these two stages. Snowball sampling was employed based on our professional contacts from colleagues, friends, and relatives who recommended managers/directors in tourism firms they knew personally. Therefore, unused questionnaires from convenience sampling were used for the snowball samples. After a respondent completed a questionnaire, we asked if they knew someone who could answer the questionnaire like what they had done. This tactic worked well, and 125 responses were generated using this sampling technique.

In total, 247 responses were received using all three sampling techniques. A paper questionnaire was primarily used unless participants requested an online version for their convenience, in which case they were sent the same version created on Google. This approach ensured the questionnaire was sent to the right respondents and not to online groups and forums where we could not control the number of questionnaires distributed and thus calculate the response rate. Overall, 247 responses to 1,000 disseminated questionnaires meant the response rate was 24.7%.

To ensure the integrity of the research and the voluntary nature of the participation, we did not give financial incentives to the respondents to respond to the survey. After deleting three responses with a missing (incomplete) rate of more than 30% and two unengaged responses, 242 were deemed usable. This represented a response rate of 24.2%, within the acceptable response rate (13.8–56.2%) for academic research on organizations (Baruch and Holtom 2008).

Twelve items from the original scales adapted from previous studies were removed in this research because of low factor loadings (less than 0.6), of large standardized residual covariance (larger than |2.58|) (Byrne 2010, p. 77), and the high correlation with other indicators simultaneously. Details of the removed items and reasons for such removal are in Supplementary Information 1.

The questionnaire was first prepared in English. To ensure the measures in both the source (English) and target (Vietnamese) versions were similar, standard translation and backward translation procedures were applied (Brislin 1970). The final English version was translated into Vietnamese (by a qualified translator). The Vietnamese version was then back-translated by other suitably qualified and experienced researchers working in strategic management and tourism management disciplines. Most of the corrections made were related to enhancing the explanations of items and further clarification in the introduction to the questionnaire. After correcting and clarifying the changes, another Vietnamese version was reviewed and refined to modify the wording to make it less technical and more understandable to potential Vietnamese respondents. The questionnaire was then sent to nine senior Vietnamese researchers in the UK and Vietnam to check whether further changes were required.

Before commencing the data collection, the questionnaire was sent to three experts working in the tourism sector to assess whether there were any possible misunderstandings and lack of coherence in the terminologies used in the industry. Each of these experts had over 15 years of experience working in the tourism sector and tourism training. They suggested that greater clarification was required to make the questionnaire more specific to the tourism industry. For example, it was agreed that ‘service’ should be clarified as a ‘tourism service’ so that respondents could quickly grasp the question's meaning. The final questions used for a survey are presented in Appendix 6..

General information on the organizations studied is presented in Table 1.

We validated the measurement models by performing CFA for six constructs. Items were restricted to load only onto their priori specified factor and were allowed to correlate with one another. We refined the measurement model by removing indicators with factor loadings lower than 0.6 and then re-running the CFA.² A summary of the average variance extracted and the construct reliabilities of the final measurement model is presented in Table 2. The overall fitness indices suggest a good fit for the measurement model (χ² = 569.869; df = 383; p = 0.000; χ²/df = 1.558; CFI = 0.94; TLI = 0.93; RMSEA = 0.048). Each item significantly loads onto its respective construct (p < 0.001) with values ranging from 0.64 to 0.89. Each construct has high composite reliability (ranging from 0.76 to 0.91), exceeding the usual 0.70 benchmarks (Hair et al. 2014). Convergent validity is satisfactory as the standardized loading for each item and the average variance extracted (AVE) both exceed the 0.5 thresholds recommended by (Hair et al. 2014). The internal consistency of the multi-item scales is also satisfactory as the composite reliability (CR) of each exceeds the 0.7 cut-off recommended by Hair et al. (2014). The details of the CFA model and results can be found in Supplementary Information 2.

Table 3 shows the correlation matrix and means for all constructs of the model.

We then examined the possibility of common methods bias following Podsakoff et al. (2003) by performing Harman’s one-factor test in two steps (Aguirre-Urreta and Hu 2019). First, all the variables were entered into an exploratory factor analysis. As a result, no single factor emerged that accounted for the majority of the variance (ranging from 28.63 to 50%). We, therefore, concluded that there is no common factor. Second the standardized regression weights (β) of the two models–with and without Common Latent Factor – were compared. If the differences between the two coefficients had been larger than 0.2, this might have indicated common method bias. However, the results suggested that common method bias was not a concern and was unlikely to confound the interpretations of our results. The matrix that indicates discriminant validity is presented in Table 4.

Table 5 presents the path analysis results using SEM on the whole sample. This shows our baseline model was a good fit (χ² = 784.167; df = 393; p = 0.000; χ²/df = 1.995; CFI = 0.89; TLI = 0.88; RMSEA = 0.064).

The Chi-squared results (χ²/df = 1.995 < 3) and RMSEA (< 0.07) show a good fit. However, CFI and TLI were slightly below the suggested level of at least 0.92 with N < 250 and m ≥ 30 (Hair et al. 2014, p. 584), suggesting a mediocre fit. Theoretical support is required to specify the model and achieve an ideal fit. Furthermore, it is important that the model specifications best approximate the theory to be tested rather than increase model fit (Hair et al. 2014). Byrne (2010) also emphasizes that fit indexes do not reflect the plausibility of a model and that judgments do depend on researchers. We adopt Mai et al.’s (2021, p. 11) “best-to-fit-a-specific purpose thinking” while testing and evaluating the models to determine the best performing fit indicators and cut-off values.

As shown in Table 5, within the model the positive impacts of Human Capital (β = 0.354; p = 0.000), Organizational Learning (β = 0.404; p = 0.000), Environmental Dynamism (β = 0.447; p = 0.000), and Digital Marketing (β = 0.188; p = 0.017 < 0.05) were supported. Therefore, H1, H2, H3, and H4 are confirmed. Regarding the antecedents that influence Dynamic Capabilities, Environmental Dynamism has the most substantial influence (β = 0.447), followed by Organizational Learning (β = 0.404), Human Capital (β = 0.354), and Digital Marketing (β = 0.188).

The full results of the baseline model are displayed in Fig. 2 below.

For robustness check, we conducted a model variation test by selecting 200 cases out of the overall sample of 242. SEM was applied to a sub-sample of 200 initial observations from the entire sample, the results of which are reported in Table 5. The fit indices of the robust check model are also satisfactory (χ² = 736.219; df = 393; p = 0.000; χ²/df = 1.873; CFI = 0.882; TLI = 0.869; RMSEA = 0.066). The results presented in Table 6 are consistent with those reported in Table 5, suggesting that our findings are robust.

Alternative models were built by adding further direct paths between variables to assess whether each alternative model was better than the hypothesized model and whether the hypothesis test results held. Four additional paths were added, and the new models were run. The results are presented in Table 7.

Comparing the fitness of the alternative models with the baseline model (χ² = 784.167; df = 393; p = 0.000; χ²/df = 1.995; CFI = 0.89; TLI = 0.88; RMSEA = 0.064), all the fitness indexes are somewhat equal to or below those of the baseline model and generate only a minimal difference. Hence, these additional paths and alternative models were not selected. Accordingly, it can be concluded that the current research model was the most appropriate and the hypothesis test results were robust.

Nevertheless, the alternative models still passed the acceptable threshold for the fitness indices and indicated the potential mediation effect of dynamic capabilities. We subsequently conducted further tests to determine whether dynamic capabilities mediate the relationships between antecedents (Human Capital, Organizational Learning, Environmental Dynamism, Digital Marketing) and the outcome (Competitive Advantage). Following Zhao et al.'s (2010) procedure, we used bootstrapping³ (with 5,000 resamples and 95% confidence intervals) of the direct and indirect effects in AMOS 21. The bootstrapping procedure provided associated p-values for each path.

The direct and total effects of each path are shown in Table 8. The results show that dynamic capabilities have a full mediation impact on the relationship between Digital Marketing and Competitive Advantage and direct-only non-mediation, in line with Zhao et al. (2010). Even though there is no indirect effect, this should not be viewed as a failure because a significant direct effect signifies undiscovered mediators (Zhao et al. 2010). This is because dynamic capabilities are a second-order construct that might weaken the impact of each subset of capabilities (sensing, seizing, and reconfiguring). We suggest that the mediation impact might be more apparent if each subset of dynamic capabilities is tested separately as a mediator in the relationships between antecedents and the outcome.

By successfully operationalizing the measurement model for dynamic capabilities as a second-order construct, this study addresses the challenge facing firms within the tourism context. This result contributes to the theory of dynamic capabilities and the relevant framework by Teece and Pisano (1994). Previous studies usually measure dynamic capabilities as a single-dimensional construct (Hawass 2010; Verreynne et al. 2016) even though, by nature, dynamic capabilities is a multidimensional construct (Laaksonen and Peltoniemi 2018; Nyamrunda and Freeman 2021).

Furthermore, this study has investigated the antecedents and outcomes of dynamic capability both internally and externally at the organizational level, namely, at the firm level. In contrast to existing literature that predominantly focuses on the dynamic capabilities of manufacturing and hi-technology businesses, this study draws from the dynamic capabilities view as the theoretical framework to provide new insights into which factors and how they contribute to the dynamic capabilities of a key service – tourism – at the level of the firm.

Our results demonstrate that dynamic capabilities in tourism firms are influenced primarily by environmental dynamism that occurs outside organizations, followed by the organizational learning culture, human capital, and the application of digital marketing within the organization. This significance level suggests that organizations are alert to the importance of fluctuations in the external environment that may affect their ability to sense and seize changes and reconfigure their processes, routines, and activities to keep pace with external dynamism. This finding reflects the reality in Vietnam – a dynamic market–which is evaluated as one of the world's most attractive tourist destinations (World Economic Forum 2019) and predicted to be one of the most sought-after international tourist places in post-Covid-19 pandemic.⁴ Previous research by Li and Liu (2014) shows that environmental dynamism is an important determinant of dynamic capabilities. Other studies, however, do not explicitly claim that environmental dynamism is an antecedent of dynamic capabilities.

Instead, they affirm it as an important factor that influences the relationship between dynamic capabilities and organizational performance (Castiaux 2012; Drnevich and Kriauciunas 2011), provided the right dynamic capability (either sensing, seizing, or reconfiguring) is applied appropriately in line with different levels of environmental dynamism (Li et al. 2019; Piening and Salge 2015; Teece 2007). A recent study by Maldonado-Guzmán et al. (2017) finds that the external environment is the barrier most likely to limit the innovative capabilities of SMEs in the service sector in Mexico. The current shock caused by the COVID-19 pandemic to economies is likely to radically affect the different capabilities of organizations to adapt and innovate to respond to the new ‘normal’ (Papadopoulos et al. 2020; Seetharaman 2020).

An organizational learning culture has been recognized as one of the key antecedents of dynamic capabilities for Vietnamese tourism organizations. Hung et al. (2010) find that dynamic capabilities mediate the influence of organizational learning culture on organizational performance and that the organizational learning culture impacts dynamic capabilities. Knowledge resources and learning mechanisms in organizations positively influence dynamic capabilities, while the learning mechanism mediates the relationship between knowledge resources and dynamic capabilities (Chien and Tsai 2012). The influence of organizational learning culture on dynamic capabilities identified in our study confirms the importance of a learning process in organizations and further enhances the scholarly discussion of the same topic in previous studies. For example, studies by Bendig et al. (2018) argue that a firm’s knowledge-based capital is part of the micro-foundation of dynamic capabilities and firm leaders indirectly influence organizational dynamic capabilities by creating individual learning conditions. The study by Linden et al. (2019) discusses the contribution of ‘knowing’ in practice to developing dynamic capabilities.

Our results confirm that the level of knowledge, skills, and abilities among employees significantly impacts the development of dynamic capabilities in Vietnamese tourism organizations. The results are consistent with Nieves and Haller's (2014) research that uses a sample of Spanish firms in the hotel industry and Kale et al.’s (2019) research about tourist accommodation establishments. Furthermore, our finding supports the strategic management literature on tourism by pointing out that micro-foundation factors (such as people) are the drivers for dynamism, advancement, progress, or improvement in organizations (Biesenthal et al. 2019; Marzo and Scarpino 2016). Thus, this research provides conclusive support for the claim by Rothaermel and Hess (2007) and more recently in the qualitative study of smaller firms in transitional economies by Nyamrunda and Freeman (2021) that investigations of the adaptation (i.e., sensing, seizing, and reconfiguring) of firms without considering human intellectual capital is inappropriate and incomplete. Our results reinforce previous theoretical work on the important role of human capital in organizations (Macher and Mowery 2009; Nieves and Haller 2014; Xing et al. 2020). As such, our results provide empirical evidence for further research on the role of micro-foundational factors (i.e., individuals) in organizations as determinants of dynamic capabilities.

The effect of digital marketing applications on dynamic capabilities is weaker than that of human capital. Prior studies suggest that technology significantly facilitates and enables a firm’s agility and adaptability (Chakravarty et al. 2013; Vogel and Güttel 2013). Our finding shows the contrast for tourism firms. It is the human factor that facilitates dynamic capabilities, regardless of which technology is used in the firm. Nevertheless, our results coincide with those of Singh and Rao (2016), Biesenthal et al. (2019), and Nyamrunda and Freeman (2021), who find that intellectual capital has a strong effect on dynamic capabilities and contributes significantly to the integration and reconfiguration of such capabilities.

This study validates the significant relationship between dynamic capabilities and an organization's competitive advantage. The results show a consistently positive impact of dynamic capabilities on competitive advantage. Previous studies find a similar linkage between dynamic capabilities and outstanding performance of organizations (Ferreira et al. 2020; Ojha et al. 2020; Ringov 2017). The outcomes of this research and the studies mentioned above collectively reinforce the ideas of Teece (2014) that dynamic capabilities do not operate alone and must be combined with effective strategizing to yield a competitive advantage. Dynamic capabilities are thus a source of competitive advantage (Salvato and Vassolo 2018) and affect the performance of organizations through the influence of different levels of environmental dynamism (Protogerou et al. 2012). Schilke’s (2014) mixed-methods study shows that dynamic capabilities are associated with a competitive advantage in moderately dynamic rather than stable or highly dynamic environments. Schilke (2014) demonstrates that the level of external dynamism influences this relationship. The positive influence dynamic capabilities exert on organizational performance forms part of the literature on organizational capabilities, demonstrating that it is a critical source of organizational performance (Wernerfelt 1984). With competitive advantage as the source of superior performance (Porter 1985), the results of this study indicate that the relationship between dynamic capabilities and organizational performance also contributes to the literature on organizational capabilities and performance. Strong dynamic capabilities must be integrated with a good strategy to achieve substantial performance (Teece 2014).

The mediation analyses confirm no mediating effects of dynamic capabilities on the direct relationship between human capital and competitive advantage, organizational learning and competitive advantage, and environmental dynamism and competitive advantage. However, as Table 8 indicates, although the indirect effects between these antecedents and competitive advantage are insignificant, the total effect of this relationship is significant. Thus, the mediation effect of dynamic capabilities exists but is too mild (as can be seen from the VAF values) to produce a significant relationship between the first three antecedents and competitive advantage. This suggests sufficient statistical power to detect the full effect but not enough to detect the effect when decomposed into its parts (Loeys et al. 2015). Taken together, the mediation effects of dynamic capabilities are too small to detect.

Our study provides useful insights to managers looking to enhance their tourism business and the development of the tourism sector. Specifically, it points to the importance of the external environment in the operations of tourism organizations. The external environmental dynamism is even more important in emerging economies like Vietnam, where the changes are intense and continuous. Additionally, the clients’ demands in a developing country for new products and services require the volumes of products and services to be delivered to change fast and often. Therefore, managers in organizations should evaluate the environment to ensure they adopt the most suitable strategy to support their company’s operations.

Moreover, managers should build a learning culture in which people are willing to share their ideas and facilitate the learning process. This action may enhance the organization’s capabilities to change and cope with new developments in the market. In conjunction with improving the quality of new recruits, creating a positive learning culture within organizations will ensure that tourism companies have better quality human capital, thus supporting their sustained success.

Furthermore, in terms of the low influence of digital marketing on dynamic capabilities found in this study, the costly investment to digitalization for both SMEs and large tourism firms should be questioned. Even though digital marketing was ranked fourth in the order of influence on dynamic capabilities, managers in tourism firms need to expend considerably more effort in improving their cyberspace presence to attract more customers. The implication for the application of digital marketing is becoming more important, given the latest movement to more digitalization in the tourism sector and changes in tourist behavior for safety because of COVID-19 (Talwar et al. 2022).

Some limitations of this study have implications for future research. First, various control variables might have been omitted that significantly influence competitive advantages. Therefore, additional internal and external variables influencing the deployment and performance of dynamic capabilities could be included. Similarly, further work is required on the strength of each capability (influenced by each determinant). Second, our data are cross-sectional and collected in a pre-defined period. This means the results might be limited in the extent to which they reflect the influence of dynamic capabilities on long-term performance and sustained competitive advantage over time. With the shocks and disruptions caused by the ongoing COVID-19 pandemic to all economies, more extensive and longitudinal research will be important to evaluate how dynamic capabilities can be sustained during a time of crisis. Third, the responses obtained might have been subjective to individuals at the point of collecting data. It is thus possible that different results would have been attained had the data been collected at a different time. Finally, due to the limited capacity of existing software that cannot calculate the R-squared of a non-recursive model like ours, this paper is incapable of providing R-squared values for the calculated models. Instead, we report the key fit indexes that have been widely suggested and applied multivariate data analysis literature (e.g., Hair et al. 2014 and Mai et al. 2021).