To test hypotheses 3a and 3b, comparing the effects of stakeholder relationships on the categories of how innovation occurs, we use multinomial logistic regression (Field 2009). This model enables us to detect the degree to which stakeholder relationships are associated with enabling innovation to be undertaken mainly by the social enterprise itself, a mix of the social enterprise with its partners, or mainly through partners. The results, as listed in Table 3, indicate that innovation is more likely to occur in conjunction with partners compared to by the social enterprise in isolation for both opportunity identification (B = 1.45, p.01) and implementation (B = 1.32, p.01) relationships. The same is true for innovation undertaken through partners compared to by the social enterprise in isolation for both opportunity identification (B = 6.79, p.01) and for implementation (B = 3.07, p.05) relationships. Finally, innovation is more likely to occur mainly through partners compared to in conjunction with partners for opportunity identification (B = 5.34, p.05), whereas there is no significant difference for implementation (B = 1.74, p[.05). Overall, the results support hypotheses 3a and 3b of the importance of building stakeholder relationships as a means to develop social innovation.