**5. Results**

#### **5.1. Non-response bias and common method bias**

The structural equation model is a multiple regression analysis, with reflective indicators that are presented as an image of the unobserved theoretical construct, representing observed variables or measures, with the objective of strengthening the relationship of influence between the constructs [38].

In this study, we performed a univariate test of significance (*t* test), to examine existing differences between respondents who answered our questionnaire quickly and those who did not. The results (*p* < 0.05) showed the absence of significant differences between the two groups of respondents. Hence, we can assure that our sample is free from non-response bias. The methods used to reduce the risk of common method bias were several. In the survey design itself, already validated in previous investigations, short and concise items were used to reduce misunderstandings. A pretest was conducted to a group of several university experts and business specialists. Similarly, following the recommendation of Podsakoff et al. [39], a distribution of items of dependent and non-consecutive independent variables was used. Finally, before assessing the relationships between dependent and independent variables, Harman's single-factor test was performed. Unrotated factor analysis using the eigenvalue-greater-thanone criterion revealed six factors, the first explaining 17.0% of the variance. This suggests that common method bias is not a serious threat to the validity of our study.

an adequate convergent validity and implying that our set of indicators represent the same

\*The variables "Risk-taking" corresponding to factor "Risk" were excluded from the measurement model due to low

values. Accordingly, values lower than 0.7 generate a low correlation and threaten the reliability of the scale.

Finally, regarding discriminant validity, this chapter presents two necessary approaches: (1) the first approach suggests that the AVE should share more variance with its assigned indicators than with any other construct (Fornell-Larcker criterion) and (2) the second approach suggests that no item should have a higher factor load with another construct than with the one, which is assigned to measure. The results shown in **Table 2** confirm the existence of discriminant

Once the measurement model is defined and validated in all its components, we will proceed and create the second-order model, following previous research (e.g. [19]), where the latent variables of the measurement model behave as constructs' measurement variables, specifically, entrepreneurial orientation (innovativeness and proactiveness) and export performance.

underlying construct [35].

**Table 1.** Measurement model.

**First-order constructs Items Factor** 

Risk-taking \* —

**Entrepreneurial orientation**

**Export performance**

**loading**

INNOV2 0.892 0.796 INNOV3 0.876 0.767 PROA1 0.844 0.712

PROA3 0.818 0.669

EP2 0.889 0.790 EP3 0.837 0.701 EP4 0.915 0.837 EP5 0.887 0.787

Key: INNOV—Innovativeness; PROA—Proactiveness; EP—Export performance.

Innovativeness INNOV1 0.813 0.661 0.827 0.896 0.742

Proactiveness PROA2 0.959 0.920 0.853 0.908 0.767

**Item loading Cronbach's α Composite** 

EP1 0.873 0.762 0.927 0.945 0.775

**reliability**

Entrepreneurial Orientation and Firm Performance http://dx.doi.org/10.5772/intechopen.72009

**AVE**

29

validity in our study.

**5.3. Evaluation of structural model**

Next, in order to analyse and interpret the PLS-SEM results, we will assess the measurement model and evaluate and test the structural model.

#### **5.2. Evaluation of measurement model**

Results from **Table 1** show that the measurement model meets all general requirements. First, all reflective items have a load higher than 0.707, which means that the reliability of individual indicators (loading2 ) is higher than 0.5. Second, all composite reliability values and Cronbach's α values are higher than 0.70, suggesting acceptable model reliability. Third, the average variance extracted (AVE) values of all constructs are higher than 0.50, indicating


Key: INNOV—Innovativeness; PROA—Proactiveness; EP—Export performance.

\*The variables "Risk-taking" corresponding to factor "Risk" were excluded from the measurement model due to low values. Accordingly, values lower than 0.7 generate a low correlation and threaten the reliability of the scale.

**Table 1.** Measurement model.

privately held and, thus, are not required to provide detailed financial information. Many SME managers are unwilling to provide correct information about their financial performance such as revenue, annual sales and return on investment. To address these problems in SME research, it is recommended using subjective measures, such as managers' perceptions, rather than objective measures in SME research [37]. Hence, perceived export performance was measured with five items, using Okpara's [18] measurement instrument, which includes profitability indicators such as growth in sales, profit, activities and operations and perfor-

The structural equation model is a multiple regression analysis, with reflective indicators that are presented as an image of the unobserved theoretical construct, representing observed variables or measures, with the objective of strengthening the relationship of influence between

In this study, we performed a univariate test of significance (*t* test), to examine existing differences between respondents who answered our questionnaire quickly and those who did not. The results (*p* < 0.05) showed the absence of significant differences between the two groups of respondents. Hence, we can assure that our sample is free from non-response bias. The methods used to reduce the risk of common method bias were several. In the survey design itself, already validated in previous investigations, short and concise items were used to reduce misunderstandings. A pretest was conducted to a group of several university experts and business specialists. Similarly, following the recommendation of Podsakoff et al. [39], a distribution of items of dependent and non-consecutive independent variables was used. Finally, before assessing the relationships between dependent and independent variables, Harman's single-factor test was performed. Unrotated factor analysis using the eigenvalue-greater-thanone criterion revealed six factors, the first explaining 17.0% of the variance. This suggests that

Next, in order to analyse and interpret the PLS-SEM results, we will assess the measurement

Results from **Table 1** show that the measurement model meets all general requirements. First, all reflective items have a load higher than 0.707, which means that the reliability of

and Cronbach's α values are higher than 0.70, suggesting acceptable model reliability. Third, the average variance extracted (AVE) values of all constructs are higher than 0.50, indicating

) is higher than 0.5. Second, all composite reliability values

common method bias is not a serious threat to the validity of our study.

model and evaluate and test the structural model.

**5.2. Evaluation of measurement model**

individual indicators (loading2

mance in general.

28 Entrepreneurship - Trends and Challenges

the constructs [38].

**5. Results**

All constructs were assessed on a five-point Likert scale.

**5.1. Non-response bias and common method bias**

an adequate convergent validity and implying that our set of indicators represent the same underlying construct [35].

Finally, regarding discriminant validity, this chapter presents two necessary approaches: (1) the first approach suggests that the AVE should share more variance with its assigned indicators than with any other construct (Fornell-Larcker criterion) and (2) the second approach suggests that no item should have a higher factor load with another construct than with the one, which is assigned to measure. The results shown in **Table 2** confirm the existence of discriminant validity in our study.

#### **5.3. Evaluation of structural model**

Once the measurement model is defined and validated in all its components, we will proceed and create the second-order model, following previous research (e.g. [19]), where the latent variables of the measurement model behave as constructs' measurement variables, specifically, entrepreneurial orientation (innovativeness and proactiveness) and export performance.


**Table 2.** Latent constructs correlation (Fornell-Larcker Criterion).

In **Tables 3** and **4**, we present the results of reliability, convergent validity and discriminant validity corresponding to the second-order model. All data confirm the strength of our model.

Next, we will follow the five steps of Hair et al. [35] in order to measure the structural model, namely, (1) collinearity assessment between constructs, (2) structural model path coefficients, (3) coefficient of determination (R2 value), (4) predictive relevance (Q2 ) and (5) the bootstrapping method. In order to obtain coefficients magnitudes, we used the path model analysis. **Figure 1** and **Table 5** summarise these results.

**6. Discussion and conclusions**

**Table 5.** Significant testing results of the structural model path coefficients.

Innovaveness

Proacveness

**Figure 1.** Results of structural model.

This chapter seeks to contribute to the development of the literature on EO as a factor that influences export performance of small firms through a robust empirical study. The central context of this research is on SMEs, as in most world economies, which constitutes the vast majority of firms in Portugal. Understanding the effects of decisions made by management in selecting strategic orientations is crucial and highly relevant to both theory and practice. Hence, this study allowed us to conclude that entrepreneurial orientation, particularly innovation and proactiveness, has a positive and significant impact on EP (H1 supported), validating previous research (e.g., [8, 31]). Portuguese textile SMEs seek to support and stimulate new ideas, experimentation and creativity that surely result in new products, services and processes. Indeed, technological innovation encompasses research and engineering efforts focused on developing new products and processes. Product innovation includes market research, design and investment on advertising and promotion. Administrative innovation is related to the development of management systems, control techniques and organisational structure. Thus, embracing innovation can generate competitive advantage and promote superior source of growth [40]. In the long run, proactive SMEs, complemented by innovative activities [17], can be market leaders in the development of new products and technologies rather than simply follow trends [8, 9], identify future customer needs, anticipate changes in demand and search new business opportunities [40]. Certainly, export firms need to continually search for new strategies and processes to obtain a better understanding of their new countries. These results can be explained by the particular characteristics of the textile sector. In this sense, each season, firms have to launch new collections (product innovations) and try to differentiate themselves from the competition (market innovations). Thus, entrepreneurial orientation has a positive and significant impact on export performance, confirming Wiklund and Shepherd [14] beliefs. Moreover, this confirms the commitment to innovation, supported by Lumpkin and Dess [17] and Miller [8], regarding the creation of new products and services, search for new opportunities and opening of new markets; and with proactiveness, firms will be able to achieve superior performance compared to competition [31].

**ENTREPRENEURIAL ORIENTATION**

**Original sample STERR** *t* **Statistics** *p* **values 2.5% 97.5% Conclusion**

EO= > EP 0.273 0.048 5.658 0.000 0.180 0.373 H1. Supported

0.913

0.736

0.273 [+]

**0.546**

Entrepreneurial Orientation and Firm Performance http://dx.doi.org/10.5772/intechopen.72009 31

**EXPORT PERFORMANCE**

Since the fundamental objective of our PLS-SEM technique is the prediction of export performance, the quality of our theoretical model will be determined by measuring the strength of each path (β), that is the relationship between entrepreneurial orientation (EO) in the predictability of the endogenous construct export performance (EP). Thus, to study our dependent variable, the value that we have to maximise is R2 . According to Hair et al. [35] and Sarstedt et al. [36], this coefficient measures the amount of construct variance that is explained by the model, where values of 0.5 are considered to be moderate and values of 0.25 are considered to be weak.

Finally, and applying the non-parametric bootstrapping test, we evaluated the significance of mediation effects. The results show significance of coefficients shown in **Figure 1**.

Results from **Table 5** indicate that EO significantly and positively influences export performance (EP), which supports H1 (β = 0.273). Hence, innovative and proactive firms achieve superior export performance.


**Table 3.** Convergence validity and reliability indexes of the second-order model.


**Table 4.** Discriminant validity index of the second-order model.

**Figure 1.** Results of structural model.

In **Tables 3** and **4**, we present the results of reliability, convergent validity and discriminant validity corresponding to the second-order model. All data confirm the strength of our model. Next, we will follow the five steps of Hair et al. [35] in order to measure the structural model, namely, (1) collinearity assessment between constructs, (2) structural model path coefficients,

ping method. In order to obtain coefficients magnitudes, we used the path model analysis.

Since the fundamental objective of our PLS-SEM technique is the prediction of export performance, the quality of our theoretical model will be determined by measuring the strength of each path (β), that is the relationship between entrepreneurial orientation (EO) in the predictability of the endogenous construct export performance (EP). Thus, to study our dependent

[36], this coefficient measures the amount of construct variance that is explained by the model, where values of 0.5 are considered to be moderate and values of 0.25 are considered to be weak. Finally, and applying the non-parametric bootstrapping test, we evaluated the significance of

Results from **Table 5** indicate that EO significantly and positively influences export performance (EP), which supports H1 (β = 0.273). Hence, innovative and proactive firms achieve

mediation effects. The results show significance of coefficients shown in **Figure 1**.

**2.** Export performance 0.513 0.861

**3.** Proactiveness 0.352 0.303 0.876

**1.** Entrepreneurial orientation 0.880

**2.** Innovativeness 0.513 0.861

**Table 3.** Convergence validity and reliability indexes of the second-order model.

**1.** Export performance 0.880

**Table 4.** Discriminant validity index of the second-order model.

value), (4) predictive relevance (Q2

) and (5) the bootstrap-

. According to Hair et al. [35] and Sarstedt et al.

**1. 2. 3.**

**1. 2.**

**1. 2. 3.**

(3) coefficient of determination (R2

30 Entrepreneurship - Trends and Challenges

superior export performance.

**Figure 1** and **Table 5** summarise these results.

**1.** Export performance 0.880

**Table 2.** Latent constructs correlation (Fornell-Larcker Criterion).

**2.** Innovativeness 0.513 0.861

**3.** Proactiveness 0.352 0.303 0.876

variable, the value that we have to maximise is R2


**Table 5.** Significant testing results of the structural model path coefficients.
