I am trying to estimate two models using 2SLRI estimator (Terza et al., 2008). I have two main variables, x1 and x2; one of which is endogenous in each model, the other one is treated as exogenous. Say, in model 1, x1 is treated as endogenous and x2 is treated as exogenous. Similarly, in model 2, x2 is treated as endogenous and x1 is treated as exogenous. In each model, the endogenous variable is treated with some instruments. There are few other exogenous variables in both models.

The result shows, in the model 1, where x1 is treated as endogenous, coefficient for x2 is insignificant (but coefficient for x1 is highly significant). Similarly, in the model where x2 is treated as endogenous, coefficient for x1 is insignificant but coefficient for x2 is highly significant.

I really puzzled why is this happening. Could anyone please shed light upon this?

Thanks in advance.