Impact of attitude towards entrepreneurship education and role models on entrepreneurial intention

In this paper, we investigate entrepreneurial intention by applying the Theory of Planned Behaviour (TPB) by Ajzen (1991). We specifically examine the role of gender on entrepreneurial education and role models or parental self-employment (PSE), by carrying out a multi-group analysis (MGA). We used a web-based questionnaire to collect information from 216 students at a Spanish university. Data are analysed with the help of Structural Equation Modelling (SEM)–Partial Least Square (PLS). We conducted a tripartite analysis on Complete, Male, and Female Models. Regarding the Complete and Male Models, all the primary hypotheses (5 in total) were accepted, compared with four for the Female Model. In this study, the primary hypotheses focus on the core variables of the TPB. We recommend the institutionalization of traineeship, elective courses, conference and workshops on entrepreneurship to boost the entrepreneurial spirit of students. Though this study has confirmed the applicability of the TPB model to entrepreneurial intention, we did not find a significant relationship between Males and Females about their entrepreneurial intentions for some relationships. However, this study suggests that the relationship between PSE and perceived behavioural control (PBC) is stronger for Males than Females Our results have implications for entrepreneurship education scholars, program evaluators, and policymakers.

Page 3 of 30 Amofah and Saladrigues Journal of Innovation and Entrepreneurship (2022) 11:36 From the foregoing, we advance some questions: What are the entrepreneurial intentions among university students? What is the relationship between PSE and Attitude Towards Entrepreneurship (ATE) and perceived behavioural control (PBC)? What is the effect of ATEE on ATE and PBC? To what extent do the relationships between Males and Females differ? Following Entrialgo and Iglesias (2017), we examine the effect of PSE and ATEE on the antecedents of the TPB and also analyse the role of gender in these relationships. Thus, the main objective of this study is to examine the role played by ATEE and PSE in fostering entrepreneurial intention among students.
To test the validity of the model, samples were drawn from students from a university in Catalonia, Spain. According to Liñán et. al. (2011b) Catalonia has a reputation for having a hard-working population, entrepreneurial spirit, and a dynamic economy.
To our best knowledge, this is a novel approach and may encourage future research in this area. A contribution of this paper is the provision of a better understanding of the role of EE and PSE and their impact on entrepreneurial intention. Moreover, the outcomes of this study could be beneficial to policymakers to understand not only the pattern of relationships among intention antecedents but also its implications for interventions and developing entrepreneurial intention. Our paper extends the studies of Trivedi (2016) by introducing Role Model or PSE and ATEE as antecedents of the TPB and the role of gender.
The remainder of the paper is structured as follows. In the next section, the literature on entrepreneurial intention in line with TPB along with the university environment and support (which we operationalize as ATEE) is outlined. This is followed by the methodology. Finally, the results of the study and their practical implications have been provided along with direction for future research and conclusion.

Theoretical framework and hypothesis development
Entrepreneurial intention and the theory of planned behaviour Bird (1988, p. 442) defined intention as 'a state of mind directing a person's attention toward a specific object (goal) or path to achieve something (means)' . According to Bae et. al. (2014) entrepreneurial intentions are the willingness to own or venture into a business. The concept of intention and its antecedents have received immerse attention in entrepreneurship research for its importance in predicting entrepreneurial behavior.
The TPB (Ajzen, 1991(Ajzen, , 2002 is perhaps one of the most popular models that has caught the attention of researchers in these contemporary times. Thus among the many models (e.g., Shapero & Sokol, 1982;Bird, 1988) used to explain entrepreneurial intentions, none have had as much impact as Ajzen's TPB (Ajzen, 1991;Krueger et al., 2000;Liñán & Chen, 2009). As of April 2020, the TPB (Ajzen, 2012) has been subject to empirical analysis in more than 4200 papers referenced in the Web of Science bibliographical database, making it one of the popular theories in the social and behavioral sciences (Bosnjak et al., 2020). They further revealed that the TPB has gained enormous attention in disciplines such as health sciences, environmental science, business and management, and educational research. In this study, the TPB is used as a basic framework to understand students' entrepreneurial intentions. The TPB model has often been used to study the intention to start a venture in a couple of research setting (Krueger, 1993;Trivedi, 2016) and it has proven that Ajzen's TPB was an appropriate research framework for Page 4 of 30 Amofah and Saladrigues Journal of Innovation and Entrepreneurship (2022) 11:36 assessing intentions in the choice of employment (Iakovleva & Kolvereid, 2009;Kolvereid, 1996). According to the TPB, human behavior is guided by three kinds of reflections, beliefs about the likely consequences of the behavior (behavioural beliefs), beliefs about the normative expectations of others (normative beliefs), and beliefs about the presence of factors that may ease or impede performance of the behavior (control beliefs) (Bosnjak et al., 2020).

Attitude towards entrepreneurship
Ajzen (1991) conceptualized attitude as the extent to which an individual has a positive or negative evaluation of the behavior in question. In the context of this paper, this refers to how a student thinks and feels about entrepreneurship. Behavioural Attitudes may be split into Affective and Instrumental. Affective attitude refers to whether an individual perceives behavior to be enjoyable or not. Instrumental attitude on the other hand, refers to whether the behavior is beneficial or harmful. The attitude towards the behavior (entrepreneurship) is an important component concerning the perception of desirability that affects entrepreneurial intention. According to Santos et al., (2016) and Liñán et. al. (2011a), ATE has a positive impact on EI.

Subjective norm (SN)
According to Ajzen (1991), the opinion of important reference groups such as parents, spouses, friends, and relatives may also influence the behavior of a person to perform or not perform certain actions. Social norms refer to the perceived social pressure from family, friends, or significant others to perform an entrepreneurial behavior (Ajzen, 1991). Social norms tend to contribute more weakly to intention (Kolvereid & Isaksen, 2006) for individuals with a strong internal inner locus of control (Ajzen, 2002) compared to those with a strong action orientation (Bagozzi, 1992). Some studies did not establish any significant direct correlation between subjective norms (SN) and EI (Krueger et al., 2000;Liñán & Chen, 2009;Santos et al., 2016). Most studies have established that subjective norms favorably affect ATE and the PBC (Entrialgo & Iglesias, 2016;Liñán & Chen, 2009;Liñán et al. 2011a;Liñán & Santos, 2007;Trivedi, 2017). Some empirical studies (Scherer et al., 1989;Matthews & Moser, 1995;Trivedi, 2016Trivedi, , 2017 have asserted that SN influence attitude and PBC and thus indirectly EI.

Perceived behavioral control
The third and most important determinant identified by Ajzen (1991) is the perceived behavioural control. PBC examines the perceived feasibility of performing behaviour and its closely related to the perception of self-efficacy (Krueger et al., 2000). PBC is the perceived easiness or difficulty of becoming an entrepreneur (Ajzen, 1991). Although some researchers have considered PBC as similar to self-efficacy, Ajzen (2002) specifies that it is a wider construct, since it encompasses and perceived controllability of the behavior. According to Santos et. al. (2016) and Liñán et. al. (2011a), PBC has a positive impact on EIs. In general, the more favorable the attitude and SN, and the greater the perceived control, the stronger should be the individual's intention to perform the behavior in question (Bosnjak et al., 2020).
Page 5 of 30 Amofah and Saladrigues Journal of Innovation and Entrepreneurship (2022) 11:36 Entrepreneurship education and support Entrepreneurship education refers to education for entrepreneurial attitudes and skills (Bae et al., 2014). It consists of 'any pedagogical program or process of education for entrepreneurial attitudes and skills (Fayolle et al., 2006, p. 702). The debate about whether entrepreneurship can be promoted through education or not persist because of inconsistencies in previous studies. Whilst some empirical studies have found a positive impact from EE (Block et al., 2013;Iakovleva & Kolvereid, 2009;Kolvereid & Moen, 1997;Souitaris et al., 2007;Valliere, 2016;Walter & Dohse, 2012), others reported a statistically insignificant or negative relationship (Oosterbeek et al., 2010;von Graevenitz et al., 2010). Bae et. al. (2014) in their meta-analysis suggested that EE is positively related to entrepreneurial intentions. According to Upton et. al. (1995), 40% of those who pursued entrepreneurship courses started their own businesses. Liñán (2008) posits that EE can nurture a student's attitudes and intentions, as well as the establishment of a new firm. Previous studies suggest that certain university support policies and practices can promote entrepreneurial activities among students, in areas such as technology transfer offices and faculty consultants (Mian, 1996); university incubators and physical resources (Mian, 1997); and university venture funds (Lerner, 2005). Entrepreneurship Education program and the entrepreneurial support provided by universities are effective ways of obtaining the requisite knowledge about entrepreneurship and motivating young people to seek an entrepreneurial career (Henderson & Robertson, 1999;Lin & Si, 2014). The impact of education and university environment on the creation of prospective entrepreneurs and the relationship between university assistance and support and the set of new businesses have gained attention in the academic circles (Trivedi 2014). Trivedi (2016) established that the university environment and support positively affect PBC. Zhang et. al. (2014) found a positive correlation between EE and entrepreneurial intention among students.

Role models/parental self-employment
Entrepreneurial family background refers to those people whose parent(s) or family member(s) is (are) involved in self-employment (Bae et al., 2014). According to Stephens (2007) parents play a major role in how their children turn out. Parents are powerful role models for children and they can influence their children's entrepreneurial intentions. Zellweger et al. (2011) argued that entrepreneurship education is less probable to promote entrepreneurial intentions of students who come from an entrepreneurial family background. According to Bae et al. (2014), EE may be less effective on entrepreneurial intentions for students from an entrepreneurial family compared to students without an entrepreneurial family background. In fact, they failed to support the hypothesis that, the positive link between entrepreneurship education and entrepreneurial intentions will be weaker in people from an entrepreneurial family background than for those who do not come from one.

The role of gender
Most studies claim that gender plays a major role in measuring entrepreneurial and self-employment career choice intentions (Verheul et al., 2012). Gender differences Page 6 of 30 Amofah and Saladrigues Journal of Innovation and Entrepreneurship (2022) 11:36 in entrepreneurship are extensively reported in the literature (Gatewood et al., 2003;Reynolds et al., 2004). The presence of a gap between males and females in entrepreneurship has long been recognized, (de Bruin et al., 2007;Díaz-García & Jiménez-Moreno, 2010;Hughes et al., 2012). Males have higher entrepreneurial intentions than females (Crant, 1996;Haus et al., 2013;Hindle et al., 2009;Wilson et al., 2004;Zhao et al., 2005). Bae et. al. (2014) failed to support the hypothesis that the positive link between EE and entrepreneurial intentions will be weaker in males than females. Gupta et. al. (2009) and Kristiansen and Indarti (2004) found no significant difference between males and females respondents on entrepreneurial intentions.

Entrepreneurial Intention
Role Models/Parental Self-employed

Fig. 1 Conceptual framework
Page 7 of 30 Amofah and Saladrigues Journal of Innovation and Entrepreneurship (2022) 11:36 From the foregoing, the following hypotheses (see Table 1) are proposed and the conceptual framework for this study is depicted in Fig. 1.

Methodology
Following Engidaw (2021), Liñán (2008) and Ndofirepi (2020), the study is developed in a single country, institution, and culture. Thus, the empirical analysis of this survey was carried out among university students in a Spanish university in the Catalonia region. We used a structured online questionnaire. Convenience sampling technique was used, because it is a popular tool in entrepreneurship research (Kolvereid, 1996;Krueger et al., 2000;Fayolle and Gailly 2015). In addition, a study by Bosma et. al. (2008) established that young graduates (25-34 years) display the highest entrepreneurial propensity. We applied the SEM-PLS technique to examine the constructs of the paper and the relationship among them.

Sample size
We used a sample size of 216, because according to Hoyle (1995), 100 to 200 respondents is usually a good starting point in conducting path modelling. In addition, Partial Least Squares (PLS) is suitable when exploratory studies are conducted and relatively small samples are used (Sánchez-Franco & Roldán, 2005).

Measurement variables
The questionnaire was divided into the following sections: demographic, independent (ATE, SN, and PBC), dependents variables (entrepreneurial intention), and Attitude Towards Entrepreneurship Education and Parental Self-employment. The study adopted the Entrepreneurial Intention Questionnaire (EIQ) proposed by Liñán and Chen (2009) to measure ATE, PBC, and SNs. Variables were tested using a five-point Likert scale from 'Strongly Agree' to Strongly Disagree. Attitude Towards Entrepreneurship Education/University environment and support scale originally developed by Kraaijenbrink et al. (2009) and revised by Trivedi (2016) was also used in this study. Eighteen items make up the ATEE Scale and are classified into two categories; General Education Support (check items 38-44 on Appendix A) and Targeted Cognitive and Non-cognitive Support (check items 27-37 on Appendix A). ATE, SN, PBC, and ATEE constructs were measured through reflective indicators. The other constructs were measured by nominal scales due to their qualitative nature: Parental Self-employed (PSE) and gender. For PSE, we asked the respondents if their mothers or fathers were entrepreneurs. It was a binary YES/NO variable. Regarding Role Models, we asked the students if, at least, one of their parents was an entrepreneur. It was a binary Yes/No variable.

Data analysis
Structural equation modeling-Partial Least Square (SEM-PLS) was used to test the proposed model which hypothesizes a relationship between entrepreneurial intention, ATE, SN, PBC, and ATEE. Hypotheses H12 to H15 were tested using multi-group analysis (MGA).

Profile of respondents
The number of respondents was 216, out of which 110 (50.9%) were males and 106 (49.1%) were females. Regarding Parental Self-employment, 110 (50.9%) of the respondents' parents were business owners, whereas 106 (49.1%), whereas 110 (50.9%) reported on the contrary. About 97.4% of the respondents were undergraduate students, 88.2% of whom were not in employment. The majority of the students fall within 20-24 ages (71.8%) category.

PLS-SEM results
In this section, we present the results of the PLS-SEM analysis. According to Hair et al. (2010), a two-dimensional method can be applied for Structural Equation Modelling (SEM); first, a measurement model analysis and second, a structural model analysis. This two-step process guarantees scale validity and reliability.

Measurement model assessment
According to Roldán and Sanchez-Franco (2012), the first stage of the measurement model assessment consists of observing the indicator loading values of the model (in our case, the three models: Complete, Male-M, and the Female-F). Table 2 depicts the parameters. It can be seen that Composite reliability, Cronbach's alpha, and Average Variance Extracted (AVE) exceed 0.7, 0.7, and 0.5, respectively, hence meeting the recommended values in literature (Fornell & Larcker, 1981). Though reliability analysis may be conducted using item loadings of above 0.707, Sánchez-Franco and Roldán (2005) opined that for newly developed measures, a lower threshold of 0.6 may be accepted. In general, the measurement model of this study was investigated following four criteria's, i.e., (a) Item reliability, (b) Internal consistency, (c) Convergent validity, and (d) Discriminant validity. As shown in Table 2, almost all the values support the convergent validity of the composite scales for the Male and Female models, but fully for the Complete model. Prior to this, the analysis of the measurement model for the full sample found low loadings (check Appendix A) for some items and were removed, and the PLS algorithm was run again. Scores regarding item reliability, construct reliability and convergent, and discriminant validity is satisfactory (see Tables 2  and 3 Table 4. The coefficient of determination for Males and Females is also shown in Table 4. According to Höck and Ringle (2006) results above the cutoffs 0.67, 0.33, and 0.19 are 'substantial' , 'moderate' , and 'weak' , respectively. Thus the results for the three models are 'substantial' . These findings are consistent with the study by Trivedi (2016) who found 69% of the variance in the explanation of entrepreneurial intention.

Structural model analysis
Using a two-tailed t test with a significance level of 5%, the path coefficient is significant if the T-statistics is larger than 1.96. Regarding the Complete model, it can be observed that three out of the nine relationships are not significant as depicted in Table 5. For the Male model, five of the hypotheses are accepted and four are rejected (see Table 6). Whereas, four of the hypotheses associated with the Females are accepted and five rejected as depicted in Table 7. Figure 5 shows the variance explained (R square) in the dependent constructs and the path coefficients (b) for the complete model. Consistent with Chin (1998), bootstrapping (5000 re-samples) was used to generate standard errors and T-statistics. Bootstrap Collinearity assessment Collinearity is a potential issue in the structural model and that variance inflation factor (VIF) value of 5 or above typically indicates such a problem (Hair et al., 2011). The collinearity assessment results for the Combined Model are summarized  in Table 8. It can be observed that all VIF values are lower than 5, signifying that there is no indicative collinearity between each set of predictor variables.

Measurement invariance of composite models
Measurement invariance of composite models (MICOM) is a logically necessary step before conducting MGA. Hult et. al. (2008Hult et. al. ( , p. 1028 posit that: 'failure to establish data equivalence is a potential source of measurement error (i.e., discrepancies of what is    intended to be measured and what is actually measured), which accentuates the precision of estimators, reduces the power of statistical tests of hypothesis, and provides misleading results' . The MICOM procedure provides the method for studying the invariance before the multi-group analysis. After confirming the existence of invariance, the next is to apply the MGA, comparing the explained variance of each group. MICOM involves a threestep process: a. Configural invariance, b. Compositional invariance and c. Scalar invariance (equality of composite means and variances).
According to Garson (2016), running MICOM in SmartPLS normally automatically establishes configural invariance. Thus, since statistical output does not apply to the first step, we did not show it. However, steps 2 and 3 are discussed below. It must be noted that in running the MICOM, outer loadings that were insignificant were deleted. This accounts for the difference in the Algorithm figure for the MGA.

Compositional invariance
Compositional invariance is a test of the invariance of indicator weights for measurement (outer) paths between groups (Garson, 2016). According to Henseler et. al. (2016), if the results of MICOM's Steps 1 and 2 (but not step 3) show that there is lack of measurement invariance, partial measurement has been established. This result allows for the comparison of the standardized path coefficients across the groups by performing a multi-group analysis. If the analysis and tests on different required levels do not support full measurement invariance, applied research typically focusses on the least partial fulfillment of measurement invariance (Hair et al., 2010). A result of non-significance means that compositional invariance may be assumed. This implies the correlations are not significantly lower than 1.0, as depicted in Table 9. Compositional invariance has been fulfilled, because the Original Correlation is equal or greater than 5% quantile. Following Henseler et. al. (2016), we tested for scalar invariance in a way comparable to that explained in Step 2. Permutation p value tests for Male and Female differences in means and variances for each of the inner model constructs. As shown in Table 10, the   Page 17 of 30 Amofah and Saladrigues Journal of Innovation and Entrepreneurship (2022) 11:36 permutations p values for Mean Original Difference are significant. However, the permutations p values for the Variance original difference are all non-significant. From the forgoing, we can assume Partial invariance.

Multi-group analysis
Having established configural and compositional invariance in Steps 1 and 2, we could compare the path coefficients of Males and Females using a multi-group analysis. The MGA uses independent samples t tests to compare paths between groups (Keil et al., 2000). In this study, we divided the sample into two groups: males (110) and females (106). This section presents the results of the MGA for the two groups (Males and Females). According to Becker et. al. (2013) researchers who failed to consider this potential issue may draw incorrect conclusions. We start by first running the PLS Algorithm to determine whether the results for the group's specific model estimation differ. Using the 'Use Relative Values' , stronger path relationships have thicker lines and smaller path coefficients have thinner lines. As shown in Fig. 6, we can apply this representation to compare the results for Males and  Innovation and Entrepreneurship (2022) 11:36 Females. From the figure, we can see that the group specific PLS coefficients differ (e.g., ATE-EI, SN-ATE, and PBC-EI). Since there are differences in the group specific PLS path model estimations, we need to find out if these differences are significant by running the PLS-MGA. Figures 7, 8 and 9 show the absolute values, outer loadings, path coefficients, and the R Square values of Males, Females and Complete. The MGA report provides path coefficients separately for the Male and Female groups, along with bootstrap-estimated standard deviations, t values, and significance p values as well as confidence intervals. From  Figs. 7, 8 and 9, we can see differences in the regression weights or beta coefficients. However, to ascertain whether the differences are significant we have to apply the bootstrap t test in the output section on the confidence intervals. From Table 11, it can be seen that the path from ATE-EI, SN-ATE, and SN-PBC confidence intervals overlap. This implies that at the 0.05 significance level, there is no difference in path coefficients between Male and Female samples. Thus, the paths in the structural model (ATE-EI,  Table 12, it can be noted that there is significant relationship between PSE and PBC but no significant relationship between the other variables; hence hypotheses H13 is accepted but H12, H14 and H15 are rejected. According to H12, the relationship between PSE and ATE is stronger for men than women. However, there are no significant relationships between both groups, hence this hypothesis is rejected. According to H13, the relationship between PSE and PBC is stronger for men than women, hence this hypothesis is accepted. According to H14, 'The relationship between ATEE and ATE is stronger for Males than for Females' . From Table 11, it can be seen that the relationship is not significant for both groups, hence we reject this hypothesis. Regarding H15, the relationship between ATEE and PBC is stronger for Males than Females. However, results reveal that the relationship between the Male and Female groups was insignificant. Hence we reject this hypothesis.

F square
The f-square equation expresses how large a proportion of unexplained variance is accounted for by R 2 change (Hair et al., 2014). The effect size is assessed with a tool known as F square indicated in Table 13 and Fig. 8. Following Cohen (1988) an F square value of above 0.35 is considered large effect size; values ranging from 0.15 to 0.35 are medium effect size; values between 0.02 and 0.15 are considered small effect and values less than 0.02 are considered NO effect size. From Fig. 104, it can be observed that the   Innovation and Entrepreneurship (2022) 11:36 ATE-EI relationship is the highest, i.e., 0.724. This is followed by SN-ATE and SN-PBC, respectively (Fig. 10).

Mediation analysis
According to Aguinis et. al. (2017), mediation refers to the presence of an intermediate variable or mechanism that transmits the effect of an antecedent variable to an outcome. The framework (Fig. 1) for this study called for multiple mediation analysis. As shown in Table 14, there are three Total Indirect Effects. However, the Specific Indirect Effects were six as depicted in Table 15. Tables 14 and 15 reveal the running of the Consistent Algorithm. To identify which of the variables were significant we run the Consistent Bootstrapping. The results are found in Tables 16 and 17. As shown in Table 17 it can be seen that SN → ATE → EI and SN → PBC → EI are significant.

Discussion
The main claim of the TPB is that intention is influenced by three variables, i.e., ATE, SNs, and PBC. This exposition of the Ajzen model lays the foundation for the hypotheses which tested the validity of the model in the present paper. Specifically, we investigated the effect of gender on ATEE and Role Models by applying the TPB (1991). Though empirical studies in entrepreneurship have produced contradictory results, we proceeded to apply the TPB to examine students' entrepreneurial intention, because it is probably one of the most tried and tested theories in entrepreneurial research. We explored the extent to which PSE and EE impact entrepreneurial intentions. We formulated two categories of hypotheses; primary and secondary and conducted a tripartite analysis for Complete, Male and Female models. This study underscored ATE as one of the important determinants of our framework. The paper exhibited a strong and highly significant relationship between ATE and entrepreneurial intention. This confirms the findings of Krueger et al. (2000) and Mahfud et al. (2020) who reported that ATE has a significant direct relationship with entrepreneurial intention.
Regarding the Complete and Male Models, all the primary hypotheses were accepted. However, with the Female Model four out of the primary hypotheses were accepted. These results are in line with previous studies (Entrialgo & Iglesias, 2016;Liñán & Santos, 2007;Liñán et al., 2011a) which found that SNs have a significant positive correlation with ATE and PBC.   Innovation and Entrepreneurship (2022) 11:36 The relationship between ATEE and EI, and PSE and EI were both insignificant. Bae et. al. (2014), in their paper, reported a statistically significant but small positive relationship between entrepreneurship education and entrepreneurial intentions.
With regards to the relationship between PSE/Role Models, the results points out that having a parent who is an entrepreneur positively influence a student's PBC (for the Complete and Female models). In addition, according to BarNir et. al. (2011), this has the probability of increasing one's knowledge, mastery, or general set of ability with regard to engaging in tasks required for becoming an entrepreneur. Interestingly, there was an insignificant relationship between PSE/Role Models and PBC for the male respondents.
According to this study the relationship between PSE and PBC is stronger for Males than Females, hence H13 is accepted. According to Wilson et. al. (2004) women tend to shy away from entrepreneurial activity more frequently than men due to a lower perception of perceived self-efficacy in carrying out entrepreneurial tasks. Verheul et. al. (2003) buttress this by emphasizing that females less frequently perceive themselves as entrepreneurs.
This study fails to fully support previous studies, on how exposure to entrepreneurial education and role models impact on Males and Females. Thus hypotheses H12, H14 and H15 were not supported. We established non-significant effects for gender and parental self-employment. These results are in line with a paper by Bae et. al. (2014), when they conducted a meta-analytic review of 73 studies. The influence of ATEE on PBC was also not significant. These findings are consistent with those of Entrialgo and Iglesias (2017).
This study has confirmed the applicability of the TPB model to entrepreneurial intention and the role of gender. However, we did not find a significant relationship between Males and Females concerning their entrepreneurial intentions for H12, H14 and H15. Therefore, gender had no significance on the path coefficients. That means the gender of a student doesn't affect the relationship between ATEE and EI. The finding further revealed that gender has no influence on the relationship between attitude and intention, which was supported by Nowinski et. al. (2019) and (Jena, 2020). These results are inconsistent with those of Santos et. al. (2016) who found that Males display higher entrepreneurial intentions than Females.

Implications and direction for future research
This study has some interesting implications. First, ATE came out as the most important variable of the model and this implies that entrepreneurial attitudes may be influenced by the relevant stakeholders in academic circles. Though we did not establish a positive correlation between PSE and ATE, influential role models can support nascent entrepreneurs. We recommend the institutionalization of traineeship, elective courses, conference and workshops on entrepreneurship to boost the entrepreneurial spirit of students. In addition, policy-makers can motivate students by providing some fiscal incentives to allow individual and business angel investments in the seed stage of their entrepreneurial activities as proposed by the European Commission (2020).
Our paper extends the studies of Trivedi (2016) by introducing Role Model or Parental Self-employment as an additional antecedent and the role of gender. This study also proximately mirrors the study by Entrialgo and Iglesias (2017), though our study used a Likert scale to measure entrepreneurial education instead of a dichotomous variable. The findings also contribute to research on parental self-employment (PSE). The results indicate that role model or parental self-employment impact on PBC for the Complete and the Female models. However, there was an insignificant relationship between parental self-employed and PBC for the Male model.
Though we found no significant relationship for ATEE on EI, we suggest that educators and the relevant stakeholders focus on how to stimulate entrepreneurial intentions through education. According to Urbano and Guerrero (2013), it is expedient to expand the scope of the university from the conventional or old-fashioned mode of knowledge to an entrepreneurial ecosystem leading to the concept of an entrepreneurial university.
Notwithstanding the importance of entrepreneurship education in the development of entrepreneurial intentions, this paper revealed that ATEE has no significant impact on ATE and PBC. This probably call for early engagement of the students to expose them to entrepreneurial education (Entrialgo & Iglesias, 2017).

Limitations
In considering the generalizability of this paper, it is important to highlight some limitations. First, the respondents were sampled from a single university in Spain. It will be exciting to replicate the study with a multi-country sample to identify the dynamics of ATEE and PSE in those countries. In addition, the majority of the students were from the Faculty of Law and Business Administration, leading to skewness of the sample characteristics. Furthermore, the insufficient number of samples in the subgroups (Male and Female) has the potential of reducing the power of analysis, leading to sampling error (Hunter & Schmidt, 2004).

Conclusions
The main objective of this study is to examine the role of gender on entrepreneurial education and role models or parental self-employment, by carrying out a multi-group analysis. The paper has contributed to the existing literature on the multi-group analysis of gender on entrepreneurial intentions among university students. Although the differences between Males and Females were not significant for three of the relationships (H12, H14 and H15), the applicability of the TPB to measure entrepreneurial intentions has been supported. This paper has reinforced attitude as one of the most important variable in the study model.