Crowd wisdom allows people to share their experiences, viewpoints, and insights; crowd wisdom can be an excellent resource for entrepreneurship education by generating three dimensions: guest speakers, mentoring programs, and peer-to-peer learning ( De Souza et al., 2016; Geng et al., 2022; Lu et al., 2017; Woraphiphat & Roopsuwankun, 2023). Thus, these are just a few examples of how crowd wisdom can be used to convey entrepreneurship.

Guest Speakers - Invite successful business founders to share with students the reality about starting and sustaining a business ( Sahedan et al., 2020). The key speakers will have an opportunity to share their experiences, provide recommendations, and respond to audience questions. The contributions of guest speakers have a significant influence on crowd wisdom because they offer new ideas, sharing their knowledge and experience, and stimulate discussion among the audience ( Sahedan et al., 2020; Sjögren & Yusuf, 2021). Overall, the knowledge held by the crowd can be influenced by guest speakers through offering new information, provoking discussion, and serving as role models ( Onjewu et al., 2021; Sahedan et al., 2020; Sjögren & Yusuf, 2021). By inviting guest speakers with diverse experiences and backgrounds, organizations can not only improve their quality decision-making processes but also encourage their audience to agree with their goals and aspirations. Hence, the higher-order hypothesis is as follows: H1a:

Guest speakers, as a dimension of crowd wisdom, positively affect entrepreneurial intention.

Mentoring Programs – Students should be offered mentoring opportunities with experienced entrepreneurs who can offer appropriate advice and support during their start-up process ( Jaiswal, 2020). Mentors may contribute to the development of business ideas, connect students with the right resources, and offer guidance based on their experiences ( Jaiswal, 2020; Onjewu et al., 2021). Mentoring programs have become increasingly popular as an excellent tool to develop and nurture entrepreneurial objectives and ambitions ( DuBois et al., 2011). Similarly, research involving immigrant entrepreneurs and recent college graduates in developing countries reports that formal mentorship programs enhance entrepreneurial self-efficacy, networking and attractiveness of entrepreneurship ( Nowiński et al., 2019; Onjewu et al., 2021). These consistent findings confirm that mentoring is an effective approach to building entrepreneurial intent in diverse institutional and cultural settings. Therefore, the higher-order hypothesis is as follows: H1b:

Mentoring programs, as a dimension of crowd wisdom, positively influence entrepreneurial intention.

Peer-to-Peer Learning– Peer-to-peer learning for entrepreneurship is a collaborative learning strategy in which prospective or early-stage business owners join together to share information, insights, and knowledge about launching and operating a company ( Markussen & Røed, 2017; G. Wang et al., 2019). It entails utilizing the group’s collective knowledge, viewpoints, and peer help to improve entrepreneurial skills, problem-solving techniques, and business development tactics ( Markussen & Røed, 2017). By sharing their knowledge and personal observations, students can learn from others’ knowledge and observations and gain a more profound understanding of entrepreneurship ( Onjewu et al., 2021). Peer-to-peer learning is a type of collaborative learning in which individuals with similar knowledge or skill levels gather to discuss and share knowledge, ideas, and skills ( Markussen & Røed, 2017; Onjewu et al., 2021; G. Wang et al., 2019). This method of teaching gives much attention to interaction and engagement among all the participants, creating a dynamic learning environment. It can take place in a range of settings, including classrooms, workplaces, the Internet, and in schools ( DuBois et al., 2011; Sahedan et al., 2020). Therefore, the higher-order hypothesis is as follows: H1c:

Peer-to-peer learning, as a dimension of crowd wisdom, positively influences entrepreneurial intention.

Entrepreneurial intention has been recognized as the most proximal cognitive predictor of entrepreneurial behavior, which plays a crucial connection between individual cognition and venture creation. Intention, based on the Theory of Planned Behavior (TPB), is defined as the result of three theoretically different but interconnected determinants: attitude towards the behavior, subjective norms and the perceived behavioral control. In the entrepreneurial domain, attitude encompasses the extent to which one has a positive or negative assessment of the value of initiating a new venture; subjective norms include the perceived social expectations and normative pressures of salient referents, such as family members, peers and role models in entrepreneurship; and perceived behavioral control represents a perceived ability and self-efficacy to carry out entrepreneurial activities.

Crowd wisdom can positively impact business objectives and decision-making ( Moritz & Block, 2015; Pavlidou et al., 2020). An entrepreneur’s vision and actions may be influenced by other persons’ collective knowledge and insights ( Kitano, 2017). Entrepreneurs should carefully study and assess the crowd wisdom they get, considering the sources’ veracity, authenticity, and applicability ( Franke et al., 2009; Pavlidou et al., 2020). Crowd wisdom can be an invaluable source of information, yet to ensure it is applicable to the entrepreneurial context, it is essential to subject it to critical analysis and contextualization ( Pavlidou et al., 2020). Crowd wisdom can affect entrepreneurs’ goals by providing feedback and validation, market insights, product and process innovation, risk mitigation, resource discovery, improved decision-making, and networking ( Busse, 2019; Geng et al., 2022; Lu et al., 2017; Pavlidou et al., 2020). Thus, influence is the capacity to persuade or convince others to behave or act in a specific way, while impact is the consequence, effect or ramifications resulting from an event ( Franke et al., 2009; Moritz & Block, 2015; Onjewu et al., 2021; Pavlidou et al., 2020; Piller & Walcher, 2006). Therefore, the following central hypothesis is developed from the above literature, which has identified the variables that are: Hypothesis H1:

Crowd wisdom and entrepreneurs’ intentions to launch a business have a strong positive relationship.

Following the literature review, a deductive approach was used to establish the conceptual framework ( Figure 1) linking the two variables. The predictor variable is represented by crowd wisdom, measured by three dimensions (guest speakers, mentoring programs, and peer-to-peer learning). The second variable is the entrepreneurs’ intention, which counts as a response variable. However, the predictor variable is displayed in second order, and the response variable comes in first order. Therefore, the following conceptual framework is drawn from the research structural equation model (SEM) and tested using Amos software to evaluate the survey data, with results approved for statistical significance in the analysis.

Descriptive statistics briefly summarize data’s fundamental characteristics in a study, including measures that describe the sample, written descriptions, or graphic representations (charts, tables, graphs). Descriptive statistics aims to summarise data in a way that is easy to understand. Table 1 presents the distribution of people by age group, by gender, using a crosstabulation of gender and age. Three age ranges are shown in Table 1: 18–20, 21–23, and 24–26. There are other totals for each gender and age group. The findings show that the sample comprises 382 people: 41% (154) are women and 59% (228) are men. The percentage distributions are calculated based on the number of individuals in each gender group. The age distribution of the male respondents is 85 ages 18–20, 78 ages 21–23, and 65 ages 24–26. Based on these numbers, 228 men responded. Conversely, 42 of the female respondents are aged 18–20, 60 are aged 21–23 and 52 are aged 24–26. These numbers amount to 154 female interviewees.

Exploratory factor analysis (EFA)

Exploratory Factor Analysis (E.F.A.) is a statistical technique for identifying latent (unobserved) correlations between variables ( Fabrigar et al., 1999). EFA primarily aims to reduce the number of observable variables to smaller latent components ( Costello & Osborne, 2005). These variables are supposed to indicate underlying processes that explain the observed intercorrelations.

The Kaiser-Meyer-Olkin (KMO) measure, according to Table 2, is a statistical measure used to evaluate whether the data used in a factor analysis is adequate ( Costello & Osborne, 2005). It focuses on how correlated the variables of the data are and how the variables can be utilized in identifying meaningful factors. The KMO measure is a 0–1 scale. The values that are nearer to 1 are a better indication of sufficient sampling ( Costello & Osborne, 2005; Fabrigar et al., 1999). Here, the KMO measure = 0.881, which shows that the data are highly correlated and can be analyzed by factor analysis. This implies that the common variance between the variables is sufficient to yield useful factors.

Bartlett’s test of Sphericity examines whether the correlation matrix is an identity matrix, indicating no correlations between variables ( Hair et al., 2014). Suppose the p-value of the test is less than the chosen significance. In that case, the null hypothesis can be rejected, meaning that the variables are significantly correlated ( Costello & Osborne, 2005; Hair et al., 2014). In this case, the p-value of Bartlett’s test is <0.001, which is less than 0.05. Therefore, we can reject the null hypothesis and conclude that the variables are significantly correlated. It supports factor analysis to extract meaningful significance from the data.

The rotated component matrix of Principal Component Analysis with Varimax rotation is shown in Table 3. Four factors were identified, matching the four constructs: Guest Speakers (GS), Mentoring Programs (MP), Peer-to-Peer Learning (PPL), and Entrepreneurial Intention (EI). For the Guest Speaker factor, two items (GS3 and GS5) showed the highest loadings (0.84 each), indicating a good fit. For the Mentoring Programs factor, the highest items were MP2 (0.89) and MP3 (0.85). Peer-to-Peer Learning saw its highest loadings on PPL3, PPL4, and PPL5 (all 0.80). Finally, the indicators for Entrepreneurial Intention were EI3 (0.86) and EI4 (0.82). All item loadings are above 0.60, and most are above 0.70, which is the accepted threshold for an item to be related to its underlying factor ( Hair et al., 2014). There were no cross-loadings, confirming factorial distinctiveness. These findings confirm the factorial validity of the scale and suggest that the four constructs are distinct, allowing the next step in reliability assessment.

Data Internal Consistency

Several statistical measures are used to evaluate internal consistency ( Hair et al., 2014). Cronbach’s alpha, which provides an extent of the inner surface of a test or scale and is stated as a value between 0 and 1, is the one that is utilized the majority of the time ( Santos, 1999). If the alpha value is more significant than 0.70, then the level of internal consistency can be regarded as good ( Hair et al., 2014; Santos, 1999).

Table 4 shows that Cronbach’s alpha exceeds 0.70, indicating that all measured variables were reliable and further analysis, including testing the measurement model through confirmatory factor analysis, is possible. Hence, the discussion in the next section is based on it. Convergent validity was measured by Average Variance Extracted (AVE), all constructs yielded an AVE of above 0.50 (Guest Speakers: AVE = 0.61; Mentoring Programs: AVE = 0.64; Peer-to-Peer Learning: AVE = 0.60; Entrepreneurial Intention: AVE = 0.58), The Fornell-Larcker criterion was used to determine discriminant validity, with the square root of the AVE of each construct, larger than the correlations between it and the others. These results confirm the convergent and discriminant validity of the measurement model, indicating that the structural model should be tested.

Confirmatory factor analysis (CFA)

This is a statistical method to test a measurement model of a construct in CB-SEM (Covariance-Based Structural Equation Modeling) ( Hair et al., 2014). CB-SEM is a statistical approach used to determine the relationships between latent and measured variables (unobserved constructs) ( Iqbal & Mia, 2021). Model fit evaluation in CB-SEM assesses how well the CFA model fits the observed data. Popular fit indices include the chi-square test, the comparative fit index (CFI), the Tucker-Lewis index (TLI), the root mean square error of approximation (R.M.S.E.A.), and the standardized root mean square residual (S.R.M.R.) ( Hair et al., 2014; Iqbal & Mia, 2021). Therefore, the appropriate indices should be interpreted to evaluate whether the model adequately matches the data.

CFA results (see Figure 2) indicate that the measurement model is well-fitting. The Goodness of Fit Index (GFI = 0.927) and Adjusted Goodness of Fit Index (AGFI = 0.906) are greater than the recommended 0.90. Comparative Fit Index (CFI = 0.969) and Tucker-Lewis Index (TLI = 0.964) exceed the recommended threshold of 0.90, indicating a good fit ( Hu and Bentler, 1999). The Root Mean Square Error of Approximation (RMSEA = 0.048) is lower than the recommended level of 0.08 ( Browne and Cudeck, 1992), and the CMIN/DF ratio (1.865) falls within the recommended range of 1–3. A combination of these indices suggests that the measurement model fits the data quite well; hence, it is reasonable to proceed to testing the structural model.

CB-SEM

When a structural equation model, a covariance analysis, is considered, it fits the data well. The CMIN/DF ratio of 1.846 is less than the recommended value of 2, indicating a reasonable model fit, and the p-value of less than 0.001 indicates that the model fits much better than a null model ( Kline & Kenny, 2005).

The goodness-of-fit measures also indicate a good model fit. The GPI of 0.927 and AGFI of 0.906 indicate an excellent fit of the model ( Hu & Bentler, 1999). The two indices exceed the recommended cutoff of 0.90 ( Hu & Bentler, 1999), and this means that the model fits well. The CFI is 0.969, and the TLI is 0.964. The RMSEA of 0.047 is also below the recommended cut-off level of 0.08, indicating that the model is useful in the investigation of the associations among the variables ( Browne & Cudeck, 1992). Thus, the results of the covariance-based structural equation model analysis indicate that the model is useful and can be applied to study the associations between the variables.

Figure 3 and Table 5 present the results of the structural equation model, which evaluated three dimensions and one central hypothesis. All three dimensions and the central hypothesis received sufficient support from the approved data. The first dimension (Guest speaker <−-- Crowd wisdom) examined whether guest speakers influenced the crowd wisdom. A regression weight estimation of 1.00 indicates a perfectly positive association between guest speakers and crowd wisdom. It led to the acceptance of the idea.

The second dimension (Mentor programs <−-- Crowd wisdom) tested the consensus-building potential of mentor programs. The critical ratio was 4.398, and the regression weight estimate was 0.289, significantly different from 1 at the 0.001 level. Therefore, the hypothesis was accepted, showing a favorable correlation between mentor programs and crowd wisdom.

The third dimension (Peer-to-peer learning <−-- Crowd wisdom) investigated their connection. At the 0.001 significance level, the estimated regression weight was 0.905, with a standard error of 0.143 and a critical ratio of 6.332. It supports the premise of a positive correlation between peer-to-peer learning and crowd wisdom.

Finally, in CB-SEM, the absolute value for the critical ratio becomes more significant than the critical value of 1.96 at a p-value of less than 0.05 ( Bentler, 1990; Hair et al., 2014; Howard, 2013; Hu & Bentler, 1999; Iqbal & Mia, 2021). The hypothesis (Entrepreneurs’ intention <−-- Crowd wisdom) yielded a critical ratio of 4.636 and a standard error of 0.064, resulting in a final regression weight estimate of 0.297. These two values had a p-value <0.001, indicating they were statistically significant. However, the central hypothesis H ₁ was shown to be statistically significant.

This study has a limitation of a small sample size since it only included undergraduate students in one university. Therefore, caution should be exercised when interpreting the results. It is recommended that a larger and more representative sample should be employed in future studies. The data is also self-reported, making it a limitation. As with any self-reports, there might be social desirability problems, and participants might not have been absolutely honest and truthful in their reporting. It would be beneficial to have other methods, including observations or physiological measurements, to supplement self-reports in future research.