According to Pearce and Robinson (2013) , the basic premise of resource-based theory (RBV) is that an organization is different from the rest, because it possesses a “unique” set of resources that it have the capacity to exploit. The RBV model suggests that the resources a business possesses are the main determinants of its productivity and success. These resources are the source of sustainable competitive advantage for businesses. Leadership, organizational structure, management systems, management practices, organizational culture… are resources that create value, and are competitive competencies.

Medalla (2005) defines BPF as reflecting the ability and efficiency of businesses in using resources such as human resources, finance and facilities to achieve business goals. Lam & Lee (2008) define performance as the outcome of a business and measured according to financial or non-financial indicators. Murphy et al. (1996) believe that BPF is a multidimensional concept. Gavrea et al. (2011) said that achievement is the expression of completing the goals and results of work; accordingly, performance is defined as the ability of a business to exploit its environment to access and use limited resources.

Currently, there are many definitions and systems for measuring BPF. BPF is a system that helps businesses plan, measure and control the results of their activities to achieve goals and create value for stakeholders (Maisel, 2001) . It is the basis for converting business strategy content into implementation conditions (Kaplan & Norton, 1993) . BPF is a set of criteria to quantify the effectiveness and efficiency of operational aspects in a business (Neely et al., 1995) .

Up to now, there have been many approaches to measuring business performance. In this study, subjective measurement approach is used to measure the performance of businesses. Subjective measures are perceptions collected from organizational members and stakeholders. Furthermore, many studies have shown that subjective measures reliably reflect objective performance outcomes (Bae & Lawler, 2000; Luo & Park, 2001; Peng & Luo, 2000) .

Dynamic capabilities are the ability of a business to integrate, build and reconfigure its internal and external capabilities to cope with rapidly changing business environments (Teece et al., 1997) . Easterby-Smith & Prieto (2008) argue that DYC are the capabilities that an organization uses, especially to cope with changes in the external environment. DYC are considered the ultimate and they provide the basis for companies to create sustainable competitive advantages (Eisenhardt & Martin, 2000) . DYC represents a type of higher-order capability that influences the speed with which a firm can respond to environmental changes (Easterby-Smith et al., 2009; Winter, 2003) . These are repetitive, patterned choices and habits that provide the ability for a business to purposefully create, expand, or modify its resource base (Helfat & Peteraf, 2009) . They include the ability to sense (Sensing), the ability to grasp (Seizing) and the ability to reconfigure (Reconfiguring) (Teece, 2007) . Sensing refers to a business’s ability to recognize opportunities in the market before its competitors. This skill can be developed by using data, analytics, and technology to scan emerging trends, uncover customer needs and identify threats from competitors. Sensing requires businesses to invest in their resources. Seizing is a business’s learning process, reflected by the ability to create internal knowledge, collect external knowledge, and assimilate internal and external knowledge through knowledge sharing to create capacity (Cepeda & Vera, 2007; Easterby-Smith et al., 2009) . Reconfiguring refers to the creation and integration of capabilities acquired internally or externally. It is the transformation of existing capabilities, that is, changing the form or appearance of existing capabilities in the enterprise (Teece, 2007) and redeploying or recombining existing capabilities (Ahuja & Katila, 2001) .

To date, there is still no unified concept of innovation. Damanpour (1996) argues that innovation is the successful implementation of unique ideas to develop an organization’s products, processes, and management systems; while Merono-Cerdan & López-Nicolas (2013) argue that innovation is the implementation of new processes in industry organizational practices, in business at work or in external relations. At the same time, Jiménez-Jiménez & Sanz-Valle (2011) said that innovation is the adaptation and application of new ideas and behaviors. Improving organizational structure and design has an impact on organizational innovation and improves internal organizational coordination and cooperation (Azar & Ciabuschi, 2017; Kraśnicka et al., 2016) . On the other hand, Alpkan et al. (2010) argue that innovation is considered an indispensable component of competitiveness that is included in the organizational structure, production process, product introduction, as well as marketing strategy in a business.

The purpose of innovation is to take advantage of the latest conditions and opportunities formed in the environment and be used to create new values and gain competitive advantage (Porter, 1985; Nonaka & Kenney, 1991; Damanpour & Schneider, 2006; Dobni, 2008; Thakur et al., 2012) . Organizations engage in innovation activities that enhance organizational value, increase profits, enhance BPF, and reduce organizational costs. In addition, innovation also aims to improve workplace satisfaction as well as labor productivity and access to knowledge resources (Vega-Jurado et al., 2009) .

Thus, innovation is a multidimensional concept with a wide scope such as strategy, organizational structure, processes and behaviors. In this study, the concept of innovation is inherited from the research of Azar & Ciabuschi (2017) . Innovation is the improvement of products/services, finding new ways of doing business, trying new ideas to improve BPF.

The original concept of ENO was proposed by Miller (1983) and Covin & Slevin (1989) , including three aspects of innovation, risk taking and initiative. This concept was later developed by Lumpkin & Dess (1996) by adding two other dimensions such as competitive aggressiveness and autonomy. Innovation is an attitude that reflects the tendency to provide support and participate in the generation of new ideas, creative processes, and changes in existing practices and technologies (Lumpkin & Dess, 1996) . Proactivity is the tendency of a person or a business to actively seek opportunities, show initiative, take action and try to create changes. Courageous risk-taking reflects the tendency to devote resources to activities or projects that have a significant likelihood of failure, but that will yield large returns if successful (Lumpkin & Dess, 1996) . Competitive aggressiveness reflects how to communicate with competitors, dare to enter the market and realize strategy faster than competitors. This is a form of business response to avoid losing competition and making marketing efforts to attract customer attention and maintain the business’s market segment. Finally, the autonomy dimension shows a significant and positive influence on business performance. This means that the role of the owner or manager is quite autonomous and essential in leading the organization to improve business performance. Additionally, autonomy causes owners or administrators to leave their comfortable positions and encourages them to look for other ideas to generate new ideas.

1) Dynamic capabilities and firm performance

There is a relationship between DYC and BPF (Teece et al., 1997; Makadok, 2001; Najib et al., 2017) . Furthermore, Mauludin et al. (2013) argue that DYC are necessary in formulating strategies in complex and rapidly changing environments, high demands for innovation, and efforts to improve organizational capabilities to overcome dynamism. DYC are organizational habits that must be acquired through very high-style learning, repetitive mastery, or repetition (Gao et al., 2014) . Based on the above background and theoretical research, the goal of this study is to examine the relationship between innovation and DYC as well as the relationship between innovation and BPF. From there, the research hypothesis is stated as:

Hypothesis H1: Dynamic capabilities have a positive impact on business performance.

Hypothesis H1a: Sensing has a positive impact on business performance.

Hypothesis H1b: Seizing has a positive impact on business performance.

Hypothesis H1c: Reconfiguration has a positive impact on business performance.

2) Dynamic capabilities and entrepreneurial orientation

Zahra et al. (2006) believe that the development of DYC is strongly associated with the ENO. The results of the study by Abdelkareem et al. (2022) shows that ENO has a meaningful impact on DYC of SMEs in Egypt. Jantunen et al. (2005) also found that ENO combined with DYC enhances the international performance of businesses in Finland. From the above foundations, the research hypothesis developed is:

Hypothesis H2: Dynamic capabilities have positive impact to entrepreneurial orientation.

3) Dynamic capabilities and innovation

One study found that innovation has a positive relationship with business performance, including profits, market share, and sales growth (Deshpande et al., 1993) . Furthermore, Craig and Dibrell (2006) demonstrated that innovation is an important requirement for business performance as well as competitiveness. Similarly, Baldwin & Johnson (1996) show that innovation has a significant effect on various measures of firm performance. From there, the research hypothesis developed is:

Hypothesis H3: Dynamic capabilities have positive impact to innovation.

4) Entrepreneurial orientation and business performance

Research results by Wahyuni and Sara (2019) and Utami & Wilopo (2018) show that ENO has a meaningful impact on the performance of SMEs. In addition, studies by (Kraus et al., 2012; Stam & Elfring, 2008; Wiklund & Shepherd, 2005) also show that ENO has a positive impact on the performance of businesses. From there, the research hypothesis was built as:

Hypothesis H4: Entrepreneurial orientation has a positive impact on business performance.

5) Innovation and business performance

Studies have found that innovation has a positive relationship with performance (Omri, 2015; Rosenbusch et al., 2011; Koellinger, 2008) . Furthermore, Craig and Dibrell (2006) demonstrated that innovation is an important requirement for the performance and competitiveness of businesses. Similarly, research by Baldwin & Johnson (1996) shows that innovation has a significant influence on various measures of BPF. Furthermore, Salavou (2002) , based on asset returns, shows that product innovation is an important determinant of BPF. Innovation helps businesses economically, creates competitive advantages and can positively affect BPF (Fallah & Lechler, 2008; Talke et al., 2011) . The positive role of innovation on performance has been supported by many empirical studies (Calantone et al., 2002; Keskin, 2006) . Therefore, the research hypothesis developed is:

Hypothesis H5: Innovation has a positive impact on business performance.

6) The mediating role of innovation and entrepreneurial orientation

Hypothesis H6: Entrepreneurial orientation plays a mediating role in the relationship between dynamic capabilities and business performance.

Hypothesis H7: Innovation plays a mediating role in the relationship between dynamic capabilities and business performance.

7) The effect of business size, type, and field on path coefficients of the structural model

Hypothesis H8a: Size of businesses moderates the path coefficients of the structural model.

Hypothesis H8b: Types of businesses moderates the path coefficients of the structural model.

Hypothesis H8c: Fields of businesses moderates the path coefficients of the structural model.

The proposed research model is presented in Figure 1 below.

The study used direct interviews using questionnaires with a 5-level Likert scale sent to leaders of tourism SMEs, including travel companies, hotels, restaurants

and tourism transportation companies in (VSCC). There were 430 surveys distributed. The results were 413 surveys, of which 8 were invalid surveys and 405 questionnaires were used. Data were processed using SmartPLS 4 software. The research sample was classified by business types, business fields, and business sizes.

The initial scale inherited the original scales including 30 observed variables, inherited from previous studies (see Table 1).

The initial scale was supplemented and completed by discussion with a group of experts including 9 people who are leaders of tourism businesses, state management officials and university lecturers.

The PLS-SEM method was performed. Assessing lower order reflective (LOC) measurement models is performed for testing: 1) reliability of indicators by outer loadings and cross loadings; 2) reliability and validity of constructs by composite reliability (CR) and average variance extracted (AVE); 3) discriminant validity by the HTMT and Fornell-Larcker indices. For the formative measurement model, the evaluation is performed for testing: a) the degree of convergence by redundancy analysis; b) the reliability of indicators by outer weights; c) the level of multicollinearity by outer multicollinearity statistics (Outer VIF); d) discriminant validity by the HTMT and Fornell-Larcker indices. Evaluation of the structural model was performed by multicollinearity testing, evaluating R², f², q², model fit, determining path coefficients, assessing direct and indirect effects, and performing multigroup analysis (MGA).

The research sample with 405 observations was collected using a convenient method combined with snow ball. In particular, the types of businesses include

Source: The authors.

sole proprietorships (Typ1) account for 30.1%, limited liability companies (Typ2) account for 45.2%, joint stock companies (Typ3) account for 11.4% and Others (Typ4) account for 13.3%. Business size with under 50 people (Siz1) accounts for 46.9%, from 50 to 100 people (Siz2) accounts for 53.1%. Regarding business fields, hotels (FLD1) account for 25.5%, restaurants (FLD2) account for 47.9%, travel businesses (FLD3) account for 10.4% and transportation businesses (FLD4) account for 16.5% (see Table 2).

The mean of the scales is high and varies from 3.868 to 4.014. The standard deviation of the scales ranges from 0.552 to 0.669. That means the scales have moderate standard deviation. This shows that there are no substantial differences between the opinions of the respondents. The mean of indicators ranges from 3.35 to 4.05 (see Tables 3(a)-(c)). The results show that the coefficient of variations (CV) of over 65% of observations is smaller than 0.15. This means the extent of variability in relation to the mean of the sample is low.

1) Assessment of reflective indicators

The results of validity assessment of the scales show that the outer loadings of all indicators are greater than 0.7, except REC1, the outer loading of which is 0.607; CEN3, the outer loading of which is 0.682; and SEZ1, the outer loading of which is 0.661. However, because the outer loading is greater than 0.4 and the AVE coefficient of the latent construct REC = 0.590 > 0.5, that of SEZ is 0.584, and of that of SEN is 0.597 > 0.5 (see Table 4), the variables CEN3, SEZ1 and

Source: The results of data processing.

REC1 are accepted (Hair et al., 2013) . This means the observed variables all converge to the basic concepts in the scales, and the scales achieve the reliability of each index (see Figure 2).

The result of collinearity statistics (VIF) shows all indicators of measurement models are smaller than 5. Thus, the models of each component have met the requirements of multicollinearity (see Table 4).

2) Construct reliability and validity

The analysis results show that the HTMT index of the all variables is equal to or less than 0.9 (see Table 5). In addition, the results of evaluating the discriminant value according to the Fornell-Larcker criteria show that the square roots of AVE are larger than the coefficients in the same column (see Table 5). Thus, the scales achieve the discriminant validity. Cross loading test results also show that the criteria are acceptable. Bootstrapping to evaluate the HTMT index was performed with a sample size of 5000. The results show that the confidence intervals of the HTMT do not contain the value 1. Thus, indicators measuring one latent construct do not measure other latent constructs.

3) Cross loadings

Cross loading test results also show that the indicators are accepted because the loadings are larger than cross loadings (see Table 6).

4) Discriminant validity

Discriminant validity was assessed using the HTMT and Fornell-Larcker criteria. The analysis results show that the HTMT index of the variables is less than 0.85 (see Table 7). In addition, the results of the evaluation of the discriminant validity according to Fornell-Larcker criteria show that the square roots of AVE

Source: The results of data processing.

Source: The results of data processing.

Source: The results of data processing.

Source: The results of data processing.

Source: The results of data processing.

are all larger than the coefficients in the same column (see Table 7). Thus, the scales achieve discriminant validity from HTMT and Fornell-Larcker criteria. Complete Bootstrapping to evaluate the HTMT index are performed with a sample size of 5000. The results show that the confidence intervals of HTMT do not contain the value 1. Thus, the scales have a high degree of discriminant validity. That is, indicators that measure one latent construct do not measure other latent constructs.

The evaluation of the construct measurement model of the latent construct DYC was performed using a two-stage embedding method. The results of redundancy analysis for the latent variable DYC show that the beta coefficient is 0.816, the R² coefficient is 0.665 and the adjusted R² is 0.665 (see Figure 3). The results of testing multicollinearity statistics (VIF) between the indicators of the formative model show that the VIF values of the indicators are all less than 5 (see Table 4).

Source: The results of data processing.

Source: The results of data processing.

The results of evaluating the statistical significance of the weights by bootstrapping with a sample size of 5000 show that all indicators for the formative model have a significance level of p value < 0.05.

The results of testing the reliability of the constructs in HOC show that the outer weights of the constructs in the formative model are greater than zero with p value < 0.05, except REC with p value = 0.401. However, outer loading of REC is 0.681 > 0.5, REC could be accepted (Hair et al., 2011) . At the same time, the outer loadings of the constructs in the reflective model are all greater than 0.7 with p value < 0.05. Thus, the scales in the HOC model meet reliability (see Table 8).

1) Collinearity statistics (VIF)

Analysis for the outer model shows that the VIF coefficients of the latent constructs and other variables in the higher-order model are all less than 5. With the

Source: The results of data processing.

inner model, the VIF values are all less than 3 (see Table 8). Thus, the models have met the requirement of multicollinearity.

2) Construct reliability and validity

The results of construct reliability assessment show that the Cronbach’s alpha of all constructs is greater than 0.7 and the composite reliability (CR) is greater than 0.7. This means that the scales have high reliability and explain ability for the research concepts in the model. Extracted variance (AVE) of all scales satisfy the condition greater than 0.5. This proves that the scales are all convergent (see Table 9).

3) Discriminant validity

Discriminant validity is assessed using the HTMT and Fornell-Larcker criteria. The analysis results show that the HTMT index of the latent variables is less than 0.85. In addition, the results of the evaluation of the discriminant validity according to Fornell-Larcker criteria show that the square roots of AVE are all larger than the coefficients in the same column. Thus, the structural model achieves discriminant validity from HTMT and Fornell-Larcker criteria (see Table 10).

Source: The results of data processing.

Source: The results of data processing.

4) Coefficient of determination (R²)

The results of data processing show that the structural model with the dependent variable BPF has an R² of 0.704 which is considered to be moderately deterministic (<0.75) and the independent variables only explain 70.4% of variability of the dependent variable. The structural model with the dependent variable ENO has an R² of 0.277. Thus, the independent variables explain 27.7% of the variation in the dependent variable, which is considered to be also moderately deterministic (>0.25 and <0.75). Model The structural model with the dependent variable INN has an R² of 0.303. Thus, the independent variables explain 30.3% of the variation in the dependent variable, which is considered to be also deterministic in the average (see Table 11).

5) Effect size of independent variables on dependent variables

Assessing the importance of the independent variables f² shows that the impact of INN on BPF, DYC on ENO as well as INN on BPF is strong (f² > 0.35). The effects of ENO on BPF and DYC on BPF are not statistically significant (see Table 12).

6) Evaluation of the predictive relevance (Q²)

The results of Bindfolding analysis show that, among the component models, the model related to DYC has no predictive relevance, with Q² equal to 0; the predictive relevance of BPF was moderate with Q² of 0.481 (<0.5) and the predictive relevance of ENO and INN was low with Q² < 0.25 (see Table 13).

7) Bootstrapping

To evaluate the significance and statistical impact of the regression coefficients, bootstrapping technique with sample number N = 5000 is used. The results of direct impact assessment, indirect impact and overall impact estimate are

Source: The results of data processing.

Source: The results of data processing.

Source: The results of data processing.

presented in Table 14.

According to the results of the structural model assessment, DYC, ENO and INN directly influence the dependent variable BPF (see Figure 4). The path coefficient of DYC -> BPF is 0.141, with p value < 0.05. DYC -> ENO has a coefficient of 0.526, with a p-value < 0.05. DYC -> INN has a coefficient of 0.550, with a p-value of 0. ENO -> BPF has a coefficient of 0.109 with a p-value < 0.05. INN -> BPF has a coefficient of 0.710, with a p-value < 0.05.

The results of the study of specific indirect effects show that DYC directly and indirectly affects BPF through both ENO and INN. The total indirect impact of

Source: The results of data processing.

DYC on BPF is 0.448. In particular, the direct transmission through ENO is 0.057 and through INN is 0.391 (see Table 15).

Results of data processing show that the SRMR coefficient of the saturated model and the estimated model are equal to 0.081 and 0.082 respectively (approximately 0.80) (Henseler et al., 2014) (see Table 16) and criteria GoF = (Average of AVE)*(Average of R²)^0.5 = 0.47 > 0.36 (Akter et al., 2011) (see Table 17). Thus, the estimated model meets the requirements of compatibility of survey data with market data.

The multigroup analysis allows us to evaluate if pre-defined data groups have

Source: The results of data processing.

Source: Results of data processing.

Source: Results of data processing.

significant differences in their group-specific parameter estimates. The results of the multigroup analysis (MGA) performed with the MICOM technique show that the path coefficient of DYC -> ENO has a difference under the impact of business with under 50 people and business with people from 50 to 100 (see Table 18).

The results of assessing the impact of business fields show that the path coefficient of DYC -> BPF, DYC -> INN, DYC -> BPF and ENO -> BPF has differences under the impact of business fields (see Tables 19-21).

The results of assessing the impact of business types show that the path coefficient of DYC -> INN has a difference under the impact of business types (see Table 22).

Source: Results of data processing.

Source: Results of data processing.

Source: Results of data processing.

Source: Results of data processing.

Source: Results of data processing.