²Stock, Yogo (2005) provide two tests that, based on the F statistic, have two purposes. The first is to test the hypothesis that in small samples the bias in the 2SLS regressions is small with respect to the endogeneity bias reported by MCO (“bias test”). The second is to use the Wald test to determine whether an instrument is considered strong, that is, that its size is close to its level for all possible configurations of the regression by instrumental variables (“size test”). Therefore, the instruments may be weak in one sense but not in another.