**Entrepreneurial Intention (EI)**

For years, there has been a classic debate about the number of options that Likert scales should incorporate [125]. However, there is consensus on the adequacy of 1–3 Likert scales [126], since it is of grea<sup>t</sup> importance that any scale includes a central point [127]. In our study, we used a 1–3 Likert scale to obtain the polarized information requested. The reliability analysis allowed us to validate the internal consistency of the scale proposed by Liñan and Chen [25]. The first step was to recode question 85, the sense of which was the reverse. However, Cronbach's alpha improved substantially when eliminating the inverse variable of this question.

**Table A9.** Variable statistics.


Cronbach's Alpha calculated after eliminating the inverse variable of question 6, an adequate scale reliability, with a value greater than 0.8.

**Table A10.** Reliability statistics.


An exploratory factor analysis was carried out with the following remaining variables, and since the variables did not present normality, following Liñan and Chen [25], a main axis factorization was applied as an extraction method. The resulting factor analysis presented an adequate goodness of fit, the level of significance of Bartlett's sphericity test is less than 0.05, and the Kaiser–Meyer–Olkin index (KMO) exceeded the minimum value of 0.50 and was very close to 1, so its significance was optimal.

**Table A11.** KMO and Bartlett's test.


There was only one factor with an eigenvalue greater than 1. As a consequence, a single factor was extracted.

**Table A12.** Total variance explained.


Extraction method: factorization of main axes.

> The factorial loads are shown in Table A13.

#### **Table A13.** Factor matrix.


Extraction method: factorization of the principal axis. One factor was extracted, and six iterations were required.

Next, the factor scores of each student for the entrepreneurial intention factor (EI) were calculated using regression. The coefficients of the factor scores were as follows:


**Table A14.** Coefficient matrix for calculating factor scores.

Extraction method: factorization of the principal axis. Rotation method: varimax normalization with Kaiser.Factor score method: regression.

#### **Corruption Perception (COPER) and Corruption Normalization (CONOR)**

The main axis factorization [25] was applied as an extraction method because the available data did not show normality. In addition, following Kaiser [128], a varimax rotation was applied, with the aim of ensuring that the factors had a few high saturations in the variables and many that were almost zero. Thus, there were factors with high correlations and with a small number of variables and with null correlations, thus leaving the variance of the factors redistributed.

The first factorial analysis presented a correct goodness of fit: The significance level of the Bartlett sphericity test was less than 0.05, and the Kaiser–Meyer–Olkin index (KMO) reached a minimum value greater than 0.50 and was very close to 1, therefore its significance is very high.

#### **Table A15.** KMO and Bartlett's test.


The correspondence of factors with corruption perception (Factor 1, COPER) and another that corresponded to corruption normalization (Factor 2, CONOR) was noted.

**Table A16.** Rotated factor matrix.


Extraction method: factorization of the principal axis. Rotation method: varimax normalization with Kaiser. The rotation converged in five iterations.

The empirical evidence showed two additional factors that were not used in this study. Consequently, the items that did not load on the two indicated factors were ignored, and the exploratory factor analysis was carried out again for questions 12–15 and 20–23.

In the second analysis, the goodness-of-fit tests were also adequate: **Table A17.** KMO and Bartlett's test.


Two factors with eigenvalues greater than 1 were reported:


Extraction method: factorization of main axes.

The factor load analysis of the rotated factor matrix showed a correspondence with the factors of corruption perception (Factor 1, COPER) and corruption normalization (Factor 2, CONOR).


**Table A19.** Rotated factor matrix.

Extraction method: factorization of the principal axis. Rotation method: varimax normalization with Kaiser. The rotation converged in three iterations.

As indicated, the factor scores of each individual were calculated for the COPER factor and for the CONOR factor by means of a regression. The coefficients for the calculation are shown in Table A20.


**Table A20.** Coefficient matrix for calculating factor scores.

Extraction method: factorization of the principal axis. Rotation method: varimax normalization with Kaiser.Factor score method: regression.

The reliability analysis allows the checking of the internal consistency of both factors. For this, Cronbach's alpha was used. As we can see, the scale was reliable, and it was not necessary to eliminate any variable in both cases. The reliability of the COPER scale was correct, since Cronbach's alpha was greater than 0.8. Cronbach's alpha did not improve when removing an item.

**Table A21.** Reliability statistics.


**Table A22.** Statistics total-item.


The reliability of the COPER scale was correct, since Cronbach's alpha was greater than 0.8. Cronbach's alpha did not improve when removing an item.

**Table A23.** Reliability statistics.


**Table A24.** Statistics total-item.


#### **Appendix C. Model Proposed, Truth Table Analysis, Quine–McCluskey Algorithm, and HHI Subsample**
