Two-temperature inferential control of the ideal and the methyl acetate double feed reactive distillation (RD) systems operated neat is evaluated using constrained dynamic matrix control (CDMC) and traditional decentralized control. The columns are operated neat (no excess of a reactant) so that the control problem is particularly challenging due to the need for stoichiometric balancing of the fresh feeds. Two basic control structures, CS1 and CS2, are studied for large through-put changes. In CS1, the two fresh feeds control a sensitive reactive and stripping tray temperatures with the reboiler duty being the through-put manipulator. In CS2, the reboiler duty controls a reactive tray temperature and the light reactant feed controls a stripping tray temperature. The heavy reactant feed acts as the through-put manipulator. Rigorous dynamic simulation results show that for the ideal RD system, significant improvement in the stripping tray temperature control and the transient deviation in the bottoms purity is observed using CDMC in both the structures. For the methyl acetate system, CDMC results in significant improvement in the control of the two tray temperatures as well as transient deviations in both the distillate and bottoms purity. Results show that the magnitude of the maximum through-put change handled is noticeably higher using CDMC. These results suggest an incentive for the application of model predictive control algorithms to RD systems.