Three-Subprocess Parallel Network DEA Model for Evaluation of Container Terminal Operational Efficiency

Container terminals comprise multiple interdependent operational areas, yet conventional data envelopment analysis models often treat terminal operations as a single production process or represent them using only broad ship-side and landside functions. This study proposes a three-subprocess parallel network data envelopment analysis model that decomposes container terminal operations into ship-side, yard-side, and gate-side subprocesses. The model explicitly incorporates the storage yard as an independent operational component and distinguishes dedicated inputs and outputs from resources and container flows shared across multiple subprocesses. An undesirable output representing vessel waiting is also incorporated through a linear data transformation. The proposed model was empirically applied to 45 terminal-year observations from nine container terminals at the Port of Busan over the period 2014–2018 and compared with an existing two-stage parallel network DEA model. The proposed model yielded a mean overall efficiency score of 0.873, compared with 0.769 under the two-stage benchmark. The mean efficiencies of the ship-side, yard-side, and gate-side subprocesses were 0.895, 0.866, and 0.803, respectively, indicating that gate-side and hinterland or inter-terminal interface operations were the most frequent sources of operational inefficiency. The contribution weights of the three subprocesses were 0.499, 0.285, and 0.216, respectively, showing that yard-side and gate-side operations jointly accounted for over half of overall terminal efficiency. Sensitivity analysis further confirmed that the efficiency scores and rankings were robust to changes in the allocation coefficients.