The current paper considers a linear regression framework with two endogenous regressors, but only one instrument that is correlated with both. I demonstrate that under reasonable conditions, some of which are testable from the data, these different sources of endogeneity act in opposing...