To gauge the extent of threshold effects, we split the sample according to pre-crisis values of NSFR and EQUITY with the help of dummy variables. In particular, we indentify banks with a NSFR below 1 and classify them by buckets of EQUITY, using 7 and 12 percent as reference limits. These values are relevant from the regulatory perspective and broadly match the quartiles of the distribution. We then re-estimate the regressions over each subsample and their combinations. As before, the estimated coefficients are transformed to convey the marginal impact of each explanatory variable on the probabilities of bank failure (Table 8). Overall, the results are consistent with the idea that liquidity and capital play a complementary role in financial stability and that threshold effects are at play. In the leftmost three columns, which are computed over the subsamples of banks with weak structural liquidity, the coefficients associated with EQUITY are three times larger than those obtained in the matching baseline regressions. Furthermore, the relationship between structural liquidity and the probability of failure reverses sign and becomes statistically insignificant for the subsample of banks with low liquidity and capital, indicating that capital shortages were critical for the failing banks in this subsample.