Results from the IV estimation are displayed in Table 9 . The first stage in Model (1) shows that all three instruments negatively predict relationship intensity, as expected. Model (2) omits the potentially endogenous variables maturity and syndicate size from the ordered pro- bit, similar to Bharath et al. (2011) . The instruments continue to negatively predict relationship intensity. The coefficients on the relationship indicators are negative and statistically significant in both models. It is apparent that the 2SLS coefficients are very large compared to the Poisson regressions. On the one hand, this is consistent with other studies in the relationship literature that have found much larger coefficients in IV regressions compared to regressions not accounting for endogeneity (e.g., Bharath et al., 2011; Berger et al., 2005 ). On the other hand, it could be a sign of instrument weakness. Using the Cragg-Donald F - statistic and the Kleibergen-Paap rk F -statistic, the table shows that Model (1) rejects instrument weakness, while Model (2) fails to reject.