Balinski's strong dual feasible bases for assignment problems are generalized to transportation problems. They are celled dual feasible super bases or trees. The super trees (not necessarily dual feasible) are also a generalization of Cunningham's (primal feasible ) strong bases. A simplex algorithm for transportation problems which generates dual feasible super bases is presented. Transportation problems with total supply (demand) $D>O$, m supply and n demand nodes are solved in at most $\mu(\mu - 1)/2 + \mu(D - \alpha - \beta - \mu + 1)$ iterations and in at most $O( \mu(m + n)(D - \alpha - \beta))$ elementary steps, where n is the maximum supply, $\beta$ is the minimum demand and $\mu = min \{m - 1, n - 1 \}$.