An estimator is consistent if MSE–> 0 as n–> infinity. Since it is desirable for the MSE to be smallest for a good estimator.
It's called "consistent" as it is consistent with what we'd expect - that is larger samples give more accurate results (ie smaller variance).