It is possible. Check out the difference between a concurrent and predictive validity study. Also, range restriction corrections can approximate the unrestricted population variance and validity coefficient.