This paper designs a Malmquist productivity index (MPI) to measure the productivity change of a decision-making unit (DMU) on a specific input/output factor over time. First, factor-specific data envelopment analysis is extended by considering input/output substitution possibilities, where partial correlation is taken as the criterion of substitutability. Factors are clustered and those which are not in the same cluster with the concerned one are excluded when calculating the factorspecific efficiency. Next, the common weights global MPI is employed, in order to simultaneously have the sound properties of consistency, circularity and comparability. Common weights are generated separately for each DMU, since only the productivities of a same DMU at different periods need to be compared in the calculation process of MPI. The case of Taiwan forests after reorganization illustrates that the proposed models can provide new insights into the productivity change of DMUs.